198 R. Miucic and S. Bai 11 NLOS 10 LOS Config. 9 Time to Collision [sec] 8 7 6 5 4 3 2 Alert Warn Inform Fig. 12 V2P application performance: comparison of configured nominal timing values and timing values achieved during the test intersection has many buildings that would act as a reflector of the signal. Thus communication and application performance will be better than achieved in this testing. Future Work To deploy the DSRC V2P application is not trivial. Yet, the arguments that make V2P attractive make the challenges worth tackling. Below, we list what the primary issues and discuss possible methods to overcome them. Positioning Accuracy Enhancement Wi-Fi ranging may be useful to improve GPS accuracy. Wi-Fi ranging uses time-of- flight information to estimate the distance between the transmitter and receiver. This technology requires that both smartphone and vehicle communicate over a wide band Wi-Fi channel in addition to DSRC. We do not cover the details here.
Cooperative Vehicle to Pedestrian Safety System 199 False Alarm Suppression The V2P system needs false alarm suppression algorithms to suppress unnecessary warnings. We implemented the following false alarm suppression techniques: • canceling the warning as soon as the driver presses the brake, • using the yaw rate in the vehicle path prediction (the path prediction is a curve that is inversely proportional to the yaw rate), • taking into account rate of change in normal distance of the pedestrian to the vehicle path (if the rate is increasing it means that the pedestrian is moving away from the intersection point and is not likely to collide with the vehicle), and • arbitrating the threats (if there are many pedestrians the vehicle system warns the driver only for the most threatening pedestrian). Future false alarm approaches may include: • driver distraction monitoring (the vehicle will suppress warning if the vehicle determines that the driver is looking in the direction of the pedestrian), • map-assisted pedestrian crossing determination (vehicle can use different levels of warnings to indicate marked crosswalk or midblock collision), and • fusion with LOS (e.g. camera, radar, or LiDAR) pedestrian detection system (for better relative position estimate). Spectrum and Channel Congestion: Potential Crash Warning Options Channel Ch172 is assigned for mainly V2V communication [19]. While other messages such as Signal Phase and Timing (SPaT) and intersection map (MAP) are expected in Ch172, most of the Ch172 communication will be BSM traffic. With potentially thousands of vehicles in the 300–400 m communication range channel congestion is an issue to consider. Potentially adding thousands more mobile devices to the V2V communication channel will only contribute to the communication congestion problem. Next, we discuss several congestion mitigation techniques. There are several possible V2P communication policies that have different effects on channel congestion: 1. Mobile devices and vehicles send and receive on Ch172, calculate collision probabilities and warn the pedestrian and driver respectively. 2. The mobile device only listens to Ch172, calculates collision probability and warns the pedestrian. The vehicle does not warn the driver because there are no pedestrian messages. (Not recommended because warning in the vehicle is more effective than warning on the phone)
200 R. Miucic and S. Bai 3. Mobile devices listen to CH172, calculate collision probabilities, warn the pedestrian, and send only a few messages informing vehicle of high collision probability. The vehicle can then act to warn the driver. All the collision detection logic resides with the mobile device. 4. A variation of (2) is to send warning messages (collision probability already calculated in mobile device) on a different continuous (V2P-dedicated) service channel e.g. CH182. 5. Yet another variation can be that the vehicle sends BSMs on CH172 and receives PSMs from the dedicated service channel CH182. Similarly the mobile device sends PSMs on CH182 and receives BSMs from CH172. 6. Another slight variation of (3) is that the V2P service channel is only activated when collision is detected by the mobile device. This solution may introduce additional delays due to processing and channel switching. The mobile device needs to advertise a V2P service to control channel Ch178, then provide information about the potential collision to the vehicle on the predefined service channel. 7. In the case of (3), (4) or (5), the mobile device sends messages on a service channel. This way the messages will not contend for channel access with V2V BSMs. This method requires the vehicle and mobile to have two radios. In the case of (1) or (5) both a vehicle and a mobile independently calculate collision probability. For all cases and especially for (1), the number of PSMs can be reduced by context sensing: the mobile sends only when the pedestrian is walking, near the road, or in dangerous situations. Congestion can be further reduced by sending PSMs with reduced frequency e.g. once a second. The affected range of the PSMs can be limited by sending PSMs with lower power, for example, 12 dBm. Conclusion In this chapter, we described advancement of the joint prototype and research effort between Honda R&D and Qualcomm Research. The research involved a DSRC based collaborative pedestrian safety system. The system consisted of pedestrian and in-vehicle components. We presented communication and application perfor- mance analysis of the prototype. Also, we discussed “What industry needs to do for V2P deployment”. Pedestrian detection is possible even in NLOS situations where the vision sensor might not be able to perform. Even though the initial stage “Inform” is challenging for the NLOS case, other warning stages are working well in both LOS and NLOS cases. In our tests, the application warned the driver thirty-one (31) times and failed to warn only once. This was due to operator error - the phone was not transmitting PSMs. We observed that DSRC achieves favorable latency performance.
Cooperative Vehicle to Pedestrian Safety System 201 References 1. World Health Organization, “Global Status Report On Road Safety 2015”, 2015, online: http://www.who.int/violence_injury_prevention/road_safety_status/2015/ GSRRS2015_Summary_EN_final2.pdf?ua=1 2. Notice of Proposed Rule Making: V2V Communications, January 2017, Available: https:/ /www.federalregister.gov/documents/2017/01/12/2016-31059/federal-motor-vehicle-safety- standards-v2v-communications 3. World Health Organization, “Pedestrian safety: a road safety manual for decision-makers and practitioners”, 2013, online: http://www.who.int/roadsafety/projects/manuals/pedestrian/en/ 4. Traffic Safety Facts – Pedestrians, DOT 812375, February 2017, Available: https:// crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/812375 5. Fatality Analysis Reporting System (FARS) Database, Available: https://www- fars.nhtsa.dot.gov 6. Traffic Safety Facts – Research Note: 2015 Motor Vehicle Crashes: Overview, DOT 812318, August 2016, Available: https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/812318 7. J. Kenney, “Dedicated short-range communications (DSRC) standards in the united states,” Proceedings of the IEEE, vol. 99, no. 7, pp. 1162–1182, July 2011. 8. U.S. Department of Transportation, “Vehicle Safety Communications Applications (VSC-A) Final Report,” U.S. Department of Transportation, Tech. Rep., 2011. 9. R. Yamaguchi, D. Ikeda, Y. Nakanishi, T. Wada, and H. Okada, “A cooperative reflect transmission scheme using road infrastructure in vehicle pedestrian communications,” in Vehicular Technology Conference, 2008. VTC 2008-Fall. IEEE 68th, Sept 2008, pp. 1–5. 10. C. Sugimoto, Y. Nakamura, and T. Hashimoto, “Development of pedestrian-to-vehicle communications system prototype for pedestrian safety using both wide-area and direct communication,” in Advanced Information Networking and Applications, 2008. AINA 2008. 22nd International Conference on, March 2008, pp. 64–69. 11. A. Al Masud, M. Mondal, and K. Ahmed, “Vehicular communication system for vehicle safety using rfid,” in Communications (MICC), 2009 IEEE 9th Malaysia International Conference on, Dec 2009, pp. 697–702. 12. X. Wu, R. Miucic, S. Yang, S. Al-Stouhi, J. Misener, S. Bai, and W. hoi Chan, “Cars talk to phones: A DSRC based: Vehicle-pedestrian safety system,” 2014, 4 manuscript submitted for publication in 2014 IEEE80th Vehicular Technology Conference. 13. K. Buchholz, “Honda works to prevent vehicle-to-pedestrian accidents,” September 2013, [Online; posted 30-September-2013]. [Online]. Available: http://articles.sae.org/12408/g 14. Savari. (2013) Mobiwave. [Online]. Available: http://www.savarinetworks.com/productsn mobiwave.html 15. Y. L. Morgan, “Managing dsrc and wave standards operations in a v2v scenario,” Interna- tional Journal of Vehicular Technology, vol. 2010, p. 18, 2010. [Online]. Available: http:// www.hindawi.com/journals/ijvt/2010/797405/cta/ 16. Mobileye 5-Series User Manual. [Online]. Available: http://www.mobileye.com/wp-content/ uploads/2012/01/UM-Series-5 BOOK ALL.pdf 17. J. C. Stutts, W. W. Hunter, and W. E. Pein, “Pedestrian crash types: 1990s update,” Transporta- tion Research Record: Journal of the Transportation Research Board, vol. 1538, p. 6, 1996. [Online]. Available: http://trb.metapress.com/content/ 18. Society of Automotive Engineers, “SAE J2735 Dedicated Short Range Communications (DSRC) Message Set Dictionary,” PA SAE, Nov. 19, 2009 19. Federal Communications Commission, “Memorandum Opinion and Order: Amendment of the Commission’s Rules Regarding Dedicated Short-Range Communication Services in the 5.850- 5.925 GHz Band (5.9 GHz Band),” FCC 06-110, online: https://apps.fcc.gov/edocs_public/ attachmatch/FCC-06-110A1.pdf
5.9 GHz Spectrum Sharing Ehsan Moradi-Pari Introduction In 1999, the Federal Communication Commission (FCC) allocated the 5.9 GHz band (75 MHz of the spectrum) for DSRC-based Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communications required for intelligent transporta- tion applications [1, 2]. The proposed communication technology employs IEEE 802.11p PHY and MAC layer models while the higher layers are based on the standards defined under IEEE 1609and SAE technical committees. This technology enables low-latency reliable ad hoc communications (broadcast mode) between vehicles as well as vehicles with infrastructure. In additions to very low latency links, this technology accommodates high-speed mobility, which makes it suitable for high-speed vehicular environment. The key enabler point for the use of this technology for automotive safety system (collision detection and avoidance system) is the fact that this technology provides low latency information exchange between the vehicles which are in communication range of each other for few seconds. In 2004, FCC further defined servicing and license rules for DSRC [3]. Figure 1 depicts the DSRC bandplan. The 5.850–5.925 GHz frequency band consists of seven 10 MHz channels (i.e., channels 172 through 184) and a reserved 5 MHz segment at the low end of the band to accommodate unforeseen future developments (this segment is often referred as the guard). Channel 178 (shown in Fig. 1) is considered to be the control channel while the remaining six channels are service channels. In 2006, FCC revised the band plan by dedicating channel 172 to V2V safety communications and channel 184 for public safety use-cases. Channel 184 allows higher power longer-range communications to make it suitable for public E. Moradi-Pari ( ) 203 West Bloomfield, MI, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 R. Miucic (ed.), Connected Vehicles, Wireless Networks, https://doi.org/10.1007/978-3-319-94785-3_8
204 E. Moradi-Pari 5.850 5.855 5.865 5.875 5.885 5.895 5.905 5.915 5.925 Reserved170 172 174 176 178 180 182 182 V2V Safety Service Service Control Service Service Public Channel Safety Channel Channel Channel Channel 175 181 Service Channel Service Channel Fig. 1 DSRC bandplan Table 1 Current DSRC channel allocation for safety and mobility transmissions Channel Usage Ch. 172 • V2V safety communication (BSM) Ch. 174 • I2V intersection safety Ch. 176 • I2V road safety (RSM) Ch. 178 • I2V safety and mobility in support of reducing congestion on Ch. 172 Ch. 180 Ch. 182 • VRU safety communications (PSM) Ch. 184 • Security management downloads Control channel • Wave service advertisements • Broadcast based I2V applications • Non BSM-based safety • Mobility applications • I2V safety and mobility • security management • Public safety safety usages [4]. In addition to the above; the bandplan permits combinations of two 10 MHz channels of 174–176 (5.865–5.885 GHz) and ch. 180–ch. 182 (5.895– 5.915 GHz) to create 20 MHz service channels. In addition to FCC channel assignment, SAE technical committee recommended allocation other DSRC channels to various safety and mobility applications [5]. Table 1 explains the current plan for safety and mobility use of the 5.850– 5.925 GHz. Wi-Fi industry has been increasingly expressing interest and asking for oppor- tunities to use 5 GHz bands understanding that the 2.4 GHz Industrial, Scientific, and Medical (ISM) band has become overloaded. In response to this interest, FCC proposed opening of the 5GHz spectrum to address the need. The communications happen on the 5GHz band are governed by U-NII rules of FCC and to be more specific, 5.850–5.925 GHz usage falls under U-NII-4 rules of FCC [6]. Figure 2 illustrates the latest proposed 5 GHz bandplan for U-NII.
5.9 GHz Spectrum Sharing 205 5.150 5.250 5.350 5.470 5.725 5.850 5.925 U-NII-1 U-NII-2A U-NII-2B U-NII-2C U-NII-3 U-NII-4 Fig. 2 5 GHz bandplan Interference Avoidance Techniques Used for Wireless Local Area Network (WLAN) Wireless network throughput is usually influenced by interference caused by multiple nodes concurrent activity. IEEE 802.11 MAC protocol utilizes different methods to maintain low interference and avoid packet collisions. These methods aim at “one transmitter at a time” on a given channel to avoid concurrency. The section below reviews two competing interference avoidance methods. Clear Channel Assessment IEEE 802.11 (Wi-Fi) implements Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA). Because of CSMA/CA, 802.11 compliant devices usually listen to/sense the medium to check the status and see if the medium is idle before transmitting. This notion is often referred to as “listen-before-talk”. Wi-Fi carrier sense implements two functions: clear channel assessment (CCA) and Network Allocation Vector (NAV). CCA provides information about the current state of the medium i.e. the medium is currently busy. NAV indicates how many future frames the medium is expected to be busy following the current frame. In order to detect the current state of the medium, CCA utilizes the following techniques: (a) Carrier sensing (CS) (b) Energy detection (ED) CS is the receiver ability to detect and decode incoming Wi-Fi preamble. Preamble constitutes the criteria when a node intends to transmit a packet and introduces the transmission. It is also used as a signal to synchronize transmission timing in network, and when detected, the channel is considered busy for the duration of time dictated by PLCP length. ED function indicates the presence of energy on the current frequency range above a certain predefined threshold (CCA threshold). It should be noted that CS is envisioned to avoid interference from other 802.11 stations while ED capability of avoiding interference with other non-802.11 devices as well as 802.11 compliant devices (using a reasonable threshold which exceeds the 802.11signal energy). Note that all the nodes in the network shall
206 E. Moradi-Pari perform CCA mechanism and as expected, different nodes might have different understanding of the network. Dynamic Frequency Selection Dynamic Frequency Selection (DFS) is a mechanism which allows the radio system operating on specific spectrum to detect, and take action to avoid interference with devices that are considered primary-use or mission-critical. DFS avoids interference by selecting a frequency that is spatially separated (proven not to interfere with primary users). Devices should be equipped with means and procedures to identify designated primary-users and stop transmissions on spatially close frequencies, which could potentially interfere with primary users. Note that, unlike CCA, primary users’ detection under DFS could be implemented by the network master detectors only and associated slave devices would be informed by the master detector about the presence of primary users. One major difference between CCA and DFS is the fact that CCA requires the station to postpone the transmission for duration of time called backoff while DFS requires the station to vacate the frequency and move the operation to an alternative channel or shut down the operation. It is worth mentioning that CCA detects presence of other users (not only the licensed users) and acts based upon the detection, while DFS detects designated primary licensed users. 5.9 GHz Band Spectrum Sharing FCC issued a Notice of Proposed Rulemaking (NPRM) on February 20, 2013, regarding the feasibility and potential use of the 5.9 GHz Dedicated Short Range Communications (DSRC) spectrum by Unlicensed National Information Infrastruc- ture (U-NII) devices [7]. The FCC NPRM proposed a new plan for allocating 195 MHz of 5 GHz spectrum aimed for unlicensed devices such as 802.11ac standard compliant Wi-Fi devices operations. In addition to this spectrum allocation proposal, to promote the use of unlicensed devices, NPRM proposed new changes to U-NII-1, U-NII-2, and U-NII- 2e bands such as making U-NII-1 available for outdoor applications [6, 8]. Figure 3 illustrates the available channels and proposed new channels. Note that unlicensed devices which are enabled to operate on new channels (red labeled) are not limited to 802.11 standard compliant devices. Table 2 below indicates the new chames and changes introduced by the FCC NPRM [6]. According to FCC Docket ET 13–49, the FCC is investigating the feasibility of sharing solution for the 5.85–5.925 GHz spectrum to be shared between DSRC and unlicensed devices such as those using 802.11-based standards and the impacts
5.9 GHz Spectrum Sharing 207 Fig. 3 Current and proposed 5 GHz channels for 802.11ac from tiger team report [6] Table 2 5 GHz U-NII band Frequency (GHZ) Original plan New plan designations comparison 5.15–5.25 U-NII-1 U-NII-1 5.25–5.35 U-NII-2 U-NII-2A 5.35–5.47 U-NII-2B 5.47–5.725 U-NII-2e U-NII-2C 5.725–5.850 U-NII-3 U-NII-3 5.85–5.925 ITS U-NII-4 associated with this approach. The primary spectrum allocation for DSRC band use was granted to the transportation community. This allocation was contingent. DSRC needed to prove that it could co-exist with the other primary users such as for military radar, satellite uplinks, indoor industrial, scientific, and medical spectrum. The primary devices, proposed by the FCC for sharing the band, use a signal based on IEEE 802.11ac and would operate in the U-NII-4 band. The DSRC and U- NII-4 (802.11ac) radio channels under consideration for coexistence are shown in Fig. 4. Note that there is currently no plan to use channel 181, which was originally proposed. Interference Types To guarantee the expected performance of DSRC communications for the use by automotive safety systems co-channel and cross-channel interferences must be investigated. The potential sources for these two interference types shall be properly understood in order to establish criteria for co-existence and band sharing. Co-channel Interference Co-channel interference becomes relevant between un-license U-NII-4 devices (e.g., 802.11ac access points, APs) and DSRC equipped vehicles that are oper- ating on the same frequency channel. This interference can severely degrade the
208 E. Moradi-Pari Safety Communications 5.855 5.85 5.865 5.875 5.885 5.895 5.905 5.915 5.925 GHz 5 GHz GHz GHz GHz GHz GHz GHz ReGsHezrved Service Service Control Service Service Public DSRC V2V Safety Channel Channel Channel Channel Channel Safety Chan 172 Chan 174 Cha 176 Chan178 Chan 180 Chan 182 Chan 184 U-NII-4 . U-NII-4 U-.NII-4 20MHz Channel 20MHz Channel 20MHz Channel U-NII-4 20MHz Channel U-NII-4 U-.NII-4 Proposed U-NII-4 40MHz Channel 40MHz Channel U-NII-4 80MHz Channel U-NII-4 160MHz Channel Fig. 4 DSRC and U-NII-4 radio channels in 5.9 GHz band performance of automotive safety applications such as intersection movement assist (IMA) in the downtown areas where the APs could be densely deployed. The effect of transmission on the same frequency band could cause significant interference when transmitter devices are close enough. Note that this interference is relevant for the overlapping portion of the spectrum, which is shown in Fig. 5. Cross-Channel Interference Extraneous power from signals and transmissions on frequencies that are not spatially separated enough (this includes adjacent channel interference) is the cause of Cross-channel interference. This is the result of emitting power from transmitters on other channels. This emitted power comes from inadequate transmitter filtering required to eliminate the signal and/or non-linear behavior of the signal processing methods being used. Figure 6 depicts this interference in the case of the non- overlapping portion of the spectrum.
5.9 GHz Spectrum Sharing 209 Safety Communications 5.855 5.85 5.865 5.875 5.885 5.895 5.905 5.915 5.925 GHz 5 GHz GHz GHz GHz GHz GHz GHz ResGeHrzved Service Service Control Service Service Public DSRC V2V Safety Channel Channel Channel Channel Channel Safety Chan 172 Chan174 Chan176 Chan178 Chan180 Chan182 Chan184 U-NII-4 U-NII-4 . 20MHz Channel 20MHz Channel U-NII-4 . Proposed U-NII-4 40MHz Channel Fig. 5 Co-channel interference scenario in 5.9 GHz band for DSRC V2V safety channel Safety Communications 5.855 5.85 5.865 5.875 5.885 5.895 5.905 5.915 5.925 GHz 5 GHz GHz GHz GHz GHz GHz GHz ReGsHezrved Service Service Control Service Service Public DSRC V2V Safety Channel Channe Channel Channel Channel Safety Chan 172 Chan 174 Cha 176 Chan178 Chan 180 Chan 182 Chan 184 U-NII-4 . U-NII-4 . 20MHz Channel 20MHz Channel U-NII-4 20MHz Channel U-NII-4 . Proposed U-NII-4 40MHz Channel U-NII-4 80MHz Channel U-NII-4 160MHz Channel Fig. 6 Cross-channel interference scenario in 5.9 GHz for upper DSRC channels
210 E. Moradi-Pari Interference Mitigation Approaches Designed for ITS Bands Two interference mitigation approaches are introduced to be candidates for sharing solution: Detect and Avoid (DAA) [9] and Re-channelization [10]. DAA concept requires no changes to DSRC technology or operation and requires unlicensed devices to avoid DSRC interference through detection of DSRC preamble in the lower 45 MHz of the DSRC spectrum. Re-channelization approach requires “safety-of-life” related DSRC activities to move to the upper 30 MHz of the DSRC band and shares the lower 45 MHz of the DSRC band with non-safety related and unlicensed devices. This section studies strength and weaknesses of these two interference mitigation proposals. In general, potential performance impact of Wi-Fi (802.11ac) emissions on DSRC in the ITS band will be reviewed. Proposal 1: Detect and Avoid (DAA) DAA is about sharing the existing DSRC channels with no additional imposed requirements to DSRC. This proposal requires no changes to DSRC technology and its operation. It requires U-NII-4 unlicensed devices to avoid interference with DSRC through detection of DSRC preamble in the lower 45 MHz of the DSRC spectrum. According to DAA proposal, “NII-4 devices that operate in the 5850–5925 MHz ITS band shall be capable of detection of ITS transmissions in 10 MHz channels between 5855 MHz and 5905 MHz”. As soon as non-licensed device detects the presence of licensed DSRC usage, non-licensed device defers the transmissions and vacates the band to avoid any interference. Figure 7 explains this technique in presence and absence of DSRC communications. DAA method implements the followings: • Detection of DSRC in 5850–5925 MHz • Detection of 802.11p preambles in 10 MHz bandwidth (imposes strict limitation on lowest power level at which the receiver can detect DSRC). • Detection of 10 MHz DSRC channels in the U-NII-4 band—if any channel is busy, then compliant devices shall defer their activities not to cause co-channel or cross channel interference • Achieves >90% detection probability within a duration of time shorter than associated slot-time • Once a 10 MHz DSRC preamble (802.11p) is detected, the frequency band from 5825 to 5925 MHz will be considered busy for compliant devices for a period of time. In addition to this, during the busy period, the DSRC channels shall be screened, and any new DSRC preamble detection will further extend the busy from the time of latest DSRC signal detection.
5.9 GHz Spectrum Sharing 211 a b Wi-Fi Network DSRC Wi-Fi DSRC detector Network detector Fig. 7 DAA proposed operation in absence of DSRC communication (a) and presence of DSRC communications (b) Legend: 20MHz 40MHz separation if 40 MHz DSRC activity detected DSRC U-NII-3 80 MHz 172 160 MHz 205 MHz separation Fig. 8 DAA proposed extended vacation This technique does not introduce/add cross channel or co-channel interferences because the compliant DAA devices defer their transmission and vacate the channel as soon as DSRC activity is detected. In addition to the protection of DSRC transmissions from harmful interferences, DAA further protects DSRC from UNII-3 communications of compliant devices (who implement this mechanism) transmit- ting on frequencies right below the ITS spectrum. Under the existing DSRC rule (explained in Fig. 1), there are possibilities for harmful cross-channel interferences from UNII-3 Wi-Fi transmissions (green channels in Fig. 3) on V2V safety-of-life communication happening in channel 172. DAA proposes extending the vacation band down to 5.825 GHz (i.e. vacating Wi-Fi channel 165). Figure 8 illustrates this notion. DAA calls for modification of 802.11ac behaviour (which could potentially impose hardware modifications) to accommodate detection of DSRC preamble on 10 MHz channels throughout the lower 45 MHz of ITS spectrum to be able to defer Wi-Fi activity as soon as DSRC signal is detected. Two conventional interference avoidance techniques of CCA and DFS (explained previously) are employed by DAA design. DAA is envisioned to implement CCA in 10 MHz bandwidth DSRC channels. Note that DAA declares channel busy as soon as DSRC signal is detected. On the other hand, when the DSRC preamble is detected the compliant devices shall cease the use of band and defer the transmission for a pre-defined period of time. In other words, it will vacate the spectrum as shown in Fig. 8.
212 E. Moradi-Pari Proposal 2: Sharing Using Modified DSRC Channelization (Re-channelization) This mechanism requires DSRC safety communications to happen on the non- overlapping portion of the spectrum (upper 30 MHz of the spectrum) and un- licensed devises are expected to share the lower 45 MHz with DSRC. The proposal recommends that ITS/DSRC use only 20 MHz channels in the overlapping portion of the spectrum (lower 40 MHz of the band 5855–5895 MHz). Unlike DAA, this proposal does not implement CCA on 10 MHz channels to detect and prioritize DSRC introduced transmissions. Note that DAA required screening of 10 MHz channels and vacation of the band as soon as detection of DSRC communications. Figure 9 depicts the proposed new band plan under re-channelization. Unlike DAA technique, re-channelization proposal does not address cross channel interference of compliant devices with DSRC transmissions compared to the existing U-NII-3 scenario (interference from green channels in Fig. 3). Re- channelization recommends DSRC to move the safety-related traffic away from U-NII-3. However, it allocates the U-NII-4 channel closer to latency sensitive safety related DSRC communications frequencies. Spatial separation from the DSRC safety related traffic is reduced as a result of removing the 5 MHz reserved segment in between. Figure 10 depicts the spectral separation under re-channelization methodology. Safety Communications 5.855 5.85 5.875 5.895 5.905 5.915 5.925 GHz 5 GHz GHz GHz GHz GHz ReGsHezrved 20 MHz Safety Safety Safety DSRC Channel 20 MHz Chan 180 Chan 182 Chan 184 Channel U-NII-4 . U-NII-4 . 20MHz Channel 20MHz Channel U-NII-4 20MHz Channel U-NII-4 U-.NII-4 Proposed U-NII-4 40MHz Channel 40MHz Channel U-NII-4 80MHz Channel U-NII-4 160MHz Channel Fig. 9 Proposed re-banding of DSRC channels by re-channelization proposal
5.9 GHz Spectrum Sharing 213 Fig. 10 Re-channelization Legend: 10MHz separation from safety DSRC DSRC 20MHz communication scenario U-NII-4 20MHz 180 182 184 40 MHz 80MHz 160 MHz In addition to spatial separation reduction, re-channelization compresses DSRC safety communications to three channels. Note that Table 2 above indicates seven DSRC channel allocations for safety and mobility transmissions. Compressing seven channels DSRC traffic to three channels increases the traffic volume on each of the three channels. Increase in DSRC traffic volume potentially introduces new sources of co-channel and cross channel interferences. One of the key aspects of sharing of the spectrum is the ability to detect the DSRC signal to be able to grant priority to DSRC traffic. Re-channelization proposes using 20 MHz channels which makes the DSRC detection more difficult. In this condition, DSRC operation on 20 MHz channel, DSRC would use similar packet structure in 20 MHz channels to Wi-Fi packets. Under this condition distinction of DSRC packets would not be as easy and straightforward as decoding preambles and requires further decoding. One possible approach might be distinguishing the DSRC packets by decoding the MAC part of the received packets. In order to decode MAC, DSRC signal has to be stronger to be detected and decoded. Note that decoding the MAC header requires high signal-to-noise ratio (SNR). On the other hand, the level of noise in the proposed 20 MHz re-banding plan would be higher than default 10 MHz channel of DSRC. Another challenge associated with detection of DSRC signal in the proposed re-banding plan for overlapping portion would be the need to detect both primary 20 MHz bandwidth channel simultaneously. The two 20 MHz proposed DSRC channels in Fig. 9 should be screened with full sensitivity simultaneously by U-NII-4 compliant devices. This is necessary in order to avoid co-channel and cross- channel interferences. Note that transmission on one channel could cause cross channel interference on DSRC transmission on the other channel. While detection of DSRC traffic is a key enabler for prioritization, the strategy of how to use the spectrum without co-channel interference is very important. DSRC uses the IEEE 802.11e quality of service (QoS) mechanism, aka EDCA, by defining eight different user priorities [11]. These user priorities map four access categories (ACs) which are shown in the Table 3. The four parameters which characterize these ACs and ultimately define the channel access protocol are as follows:
214 E. Moradi-Pari Table 3 DSRC EDCA parameters as defined in SAE J2945/1 User priority AC CWmin CWmax AIFSN TXOP limit OFDM/CCKOFDM PHY 1, 2 AC_BK 15 1023 9 0 0, 3 AC_BE 15 1023 6 0 4, 5 AC_VI 15 1023 4 0 6, 7 AC_VO 2 0 3 7 Table 4 802.11ac Parameter 802.11ac default setting parameters for best effort (AC_BE) category EDCA parameter CWmin = 15 CWmax = 1023 AIFSN = 3 • Contention Window Minimum (CWmin) • Contention Window Maximum (CWmax) • Arbitration InterFrame Space Number (AIFSN) • Transmission Opportunity (TXOP) limit: The non-DSRC U-NII-4 compliant devices (Wi-Fi) use different set of param- eters than DSRC in typical traffic protocol (AC_BE category) as explained in Table 4. Lemma 1 The expected wait time for DSRC traffic to get the chance of transmission is higher than U-NII-4 Wi-Fi. Proof Note that: E [wait time] = E [backoff time + I nterf rame_space] = E [backoff _counter ∗ Slot_time] + interf rame_space = Slot_time ∗ E [backoff ] + interf rame_space Slot time and interframe space of DSRC is higher than 802.11ac Wi-Fi, and assuming uniform backoff distribution with the same CWmin is 15 in both cases (the same expected value) we can conclude that the expected wait time for DSRC traffic to get the chance of transmission is higher than 802.11ac. The EDCA parameters could be modified in order to prioritize DSRC traffic. This can be done by assigning a longer average wait-time to U-NII-4 Wi-Fi devices. Conclusion In this chapter, conventional radio local area network interference avoidance techniques, CCA and DFS are studied. To understand the interference scenarios, sources of co-channel (when devices operate on the same frequency) and cross-
5.9 GHz Spectrum Sharing 215 channel (when devices operate on near frequencies) in the case of DSRC safety communications are reviewed. It is important to investigate the impacts of potential co-channel and cross-channel interferences to guarantee the feasibility of co- existence. Key features of two interference mitigation proposals of DAA and re- channelization are introduced and compared in this chapter. According to the comparisons, DAA looks to promise more potential for protecting DSRC safety communications. Because of the reduced spatial separation, as proposed by re- channelization, moving the safety communication to upper channels (i.e., Ch. 180, 182, 184) could be potentially problematic. In addition to reduced spatial separation, the increased in volume of DSRC traffic on the upper channels, as a result of compressing seven channels communication to three channels, could impose new sources of interference on DSRC communications. Another key aspect that needs to be taken into account is the detection of DSRC traffic on the overlapping portion of the spectrum. It is important to screen the overlapping channels with full sensitivity simultaneously not to impose interference potentials on DSRC traffic. Proper detection mechanism is a necessary piece for the prioritization of any specific communication traffic on the overlapping frequencies. If the goal were to give the priority to DSRC traffic on the overlapping frequencies, modification of EDCA parameters would be required to accommodate such prioritizations. To understand the feasibility of co-existence and band sharing solution, it is important to investigate the impacts of communication traffics on each other. DSRC is targeted to provide safety-related situation awareness. There are several metrics identified by ITS community as key performance indicators for vehicular safety applications (e.g., collision avoidance). Channel busy ratio (CBR), packet error rate (PER), and information age (IA) are proven to be good indicators for this safety- related applications. CBR indicates the percentage of the time, during which the channel sensed to be busy. PER indicates the rate of losing packets and IA shows how current the received information from a specific vehicle is. It is important to show the performance of sharing mechanism in terms of these metrics to be able to make a fair judgment about the feasibility of co-existence and spectrum sharing. References 1. Amendment of Parts 2 and 90 of the Commission’s Rules to Allocate the 5.850-5.925 GHz Band to the Mobile Service for Dedicated Short Range Communications of Intelligent Transportation Services, ET Docket No. 98-95, Report and Order, 14 FCC Rcd 18221 (1999). 2. FCC 03-324 Report and Order, Dec. 17, 2003. 3. Amendment of the Commission’s Rules Regarding Dedicated Short-Range Communication Services in the 5.850-5.925 GHz Band (5.9 GHz Band), WT Docket No. 01-90; Amendment of Parts 2 and 90 of the Commission’s Rules to Allocate the 5.850-5.925 GHz Band to the Mobile Service for Dedicated Short Range Communications of Intelligent Transportation Services, ET Docket No. 98-95, Report and Order, 19 FCC Rcd 2458 (2004) (DSRC Report and Order).
216 E. Moradi-Pari 4. FCC 06-110 Memorandum Opinion and Order, July 20, 2006. 5. Dedicated Short Range Communication (DSRC) Systems Engineering Process Guidance for SAE J2945/X Documents and Common Design Concepts, Dec, 2017. 6. IEEE 802.11-15/0347r0, Final Report of DSRC Coexistence Tiger Team at 1 (Mar. 9, 2015) (Tiger Team Final Report), https://mentor.ieee.org/802.11/dcn/15/11-15-0347-00-0reg-final- report-of-dsrc-coexistence-tiger-team-clean.pdf. 7. Federal Communications Commission, “In the Matter of Revision of Part 15 of the Commis- sion’s Rules to Permit Unlicensed National Information Infrastructure (U-NII) Devices in the 5 GHz Band,” ET Docket No. 13-49, February 20, 2013. 8. Lansford, J.; Kenney, J.B.; Ecclesine, P., “Coexistence of unlicensed devices with DSRC systems in the 5.9 GHz ITS band,” IEEE Vehicular Networking Conference (VNC), Boston, 2013, pp. 9–16, 16-18 Dec. 2013. 9. Tiger Team Final Report at 6-7. See also Cisco Systems Inc. Reply at 24-28; Letter from Mary L. Brown, Senior Director, Government Affairs, Cisco Systems, Inc. to Marlene H. Dorch, Secretary, FCC (Dec. 23, 2015). 10. Tiger Team Final Report at 7-8. See also Qualcomm Inc. Comments at 5-17 (Qualcomm Comments). 11. SAE International, “Surface Vehicle Standard – On-Board System Requirements for V2V Safety Communications,” J2945™/1, Issued 2016-03.
Efficient and High Fidelity DSRC Simulation Yaser P. Fallah and S. M. Osman Gani Introduction Connected vehicle applications use Dedicated Short-Range Communications (DSRC) technology to disseminate safety critical or traffic information [1–4]. Active safety applications are arguably the most important connected vehicle applications. Given the criticality of safety applications, it is imperative to extensively research and test the applications before deployment. However, the dynamic nature of (DSRC) communication networks and vehicular traffic, and a multitude of factors that affects each, make it prohibitively expensive, and technically infeasible to conduct field tests for all possible communication and traffic scenarios. In particular, the nature of safety applications and rarity of events is such that large scale tests with over hundreds of vehicles are usually very difficult, if not impossible. Even small scale tests of the application under all communication possibilities may not be feasible. Given the difficulties of field trials, researchers generally resort to simulation to examine and verify the performance of safety applications. Simulation efforts generally target three different aspects of DSRC based safety applications: communication network, vehicle traffic and movement, and safety algorithms. While safety algorithms can often be exactly implemented in simulators, the communication network and traffic aspects have to be modeled and simulated at lower fidelity. As a result, it becomes vital to ensure that the simulation tools are precise enough and credible. The focus of this chapter is on simulation of the DSRC based communication network. Simulation of wireless communication networks requires modeling different layers of the protocol stack, in addition to the behavior of the wireless medium. Y. P. Fallah ( ) · S. M. Osman Gani 217 Department of Electrical and Computer Engineering, University of Central Florida, Orlando, FL, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 R. Miucic (ed.), Connected Vehicles, Wireless Networks, https://doi.org/10.1007/978-3-319-94785-3_9
218 Y. P. Fallah and S. M. Osman Gani While most protocol behaviors can be exactly implemented according to the standard, some details at the lower layers have to be abstracted. Modeling of the wireless medium, i.e., channel propagation behavior, is also a challenging task and exact reconstruction of what goes on in the physical wireless environment is generally impossible. As a result, there are numerous approaches to modeling and abstracting the behavior of different layers and channel models. Some of the recent works on simulation of DSRC focused on adapting existing popular network simulators such as ns-2 and ns-3. Chen et al. in [5, 6] addressed the shortcomings of the simplistic physical layer (PHY) and medium access control (MAC) implementations of ns-2. In their work, the authors separated PHY and MAC layer functionalities in a proper way so that functionalities of the network components are placed in their respective layers, improving ns-2 simulation accuracy. The simplistic approach of carrier sensing in ns-2 was also improved by integrating a “NoiseMonitor” to keep track of all interferences at a transceiver. The work in [6] also introduced a SINR based reception criteria, that determined whether a frame is successfully received or not. In [7], the authors have extensively investigated, using their 802.11 test bed, the details of physical layer capture effect by explaining various capture scenarios. Based on that study [8, 9] proposed some modifications of the 802.11 PHY in QualNet [10]. Papanastasiou et al. [11] and Mittag et al. [12] developed a detailed PHY model (PhySim-Wifi) based on ns-3 [13] in an effort to incorporate DSRC based vehicular communication to the simulator. Instead of using a frame that is defined solely by the length field of the frame, PhySim-Wifi emulates the physical bits of the frame by transforming a randomly generated bit sequence equal to the size of the length field into complex time domain samples as per IEEE 802.11 standard. Although PhySim models a realistic Wi-Fi PHY by considering bitwise processing of the signal, the trade-offs here are the higher memory and computational requirements. Overall, simulation of wireless networks requires developing models for nodes and the wireless channel. Node models should in general take into account transmitter and receiver behavior, while channel models describe the signal loss and deformation between each pair of nodes. Therefore, we describe the node model and channel model separately in this chapter. It should be noted that the node model details will depend on how detailed the channel model is (Fig. 1). Wireless Channel Models There are several approaches in modeling the propagation behavior. In a simpler form, the channel behavior can be considered as a random loss pattern affecting PHY frames. In more practical forms, the power (propagation) loss can be modeled and derived as a function of some factors that describe the environment between each pair of nodes. For example, the distance between each pair of nodes, and the
Efficient and High Fidelity DSRC Simulation 219 Application Application Send Message Receive Message Transport + Network + Transport + Network + Data Link Layers Data Link Layers Transmitter Receiver Node Model TxPower RxPower Channel Propagation Channel Model Path Loss + Shadowing + Fading Fig. 1 Simulation steps general form of the physical environment (urban, suburban, highway, etc.) may be used to derive formulas that relate power loss to distance. Other factors such as traffic density may also be considered. In more precise forms, a ray tracing [29] scheme may be used that tries to reconstruct a more precise form of the channel and considers geometry of environment features such as buildings, vehicles, etc., to calculate reflections and multipath effects more precisely. Another aspect of channel modeling is the effect of the channel loss in time. A simpler approach is to assume that an entire frame is affected by the same propagation loss. A more detailed approach will look at smaller sections of the frame, down to individual symbols. The latter approach may not be necessary if channel properties are assumed to remain unchanged during the life of a frame, although for longer frames some adjustment may be needed. Generally, the more details a model considers, the more computationally expen- sive it becomes. As a result, many of the popular network simulators such as ns-2, ns-3, OMNET, etc., use frame level simulation of channel effect. The propagation model is also abstracted in most cases to formulas that try to recreate the impact of large scale channel loss, shadowing and fading on an entire frame. Some simulators allow for employing of ray tracing and more granular modeling of the propagation loss (e.g., OPNET allows both approaches). Nevertheless, such detailed propagation models are computationally costly. In this chapter, we detail the approach taken by ns-3 (an open source simulator widely used by academic and industry researchers) and present the corrections that we have made to ns-3 based on DSRC field tests, to derive a higher fidelity frame level simulator, which remains computationally efficient.
220 Y. P. Fallah and S. M. Osman Gani Node Model The node behavior needs to be modeled to account for all layers of communication and application. At the highest layer, the application module generates the safety messages and contains the implementation of targeted safety applications such as forward collision warning (FCW) [16, 17], intersection movement assist (IMA) etc. It also contains the implementation of channel congestion control algorithms such as J2945/1 [23] that are put in place for adapting the channel load from the application layer by scheduling the generation of basic safety message (BSM [24]) based on the channel condition feedback. Channel congestion can be regulated using various parameters, for example, message generation rate, transmit power etc. Generated messages at the application layer are passed down to the lower layers, and eventually transmitted through the communication channel. Received messages at the application layer can be used to calculate various metrics such as channel busy percentage (CBP), packet error rate (PER), vehicle density within the area of interest etc., which can be used as feedback parameters for the channel congestion control algorithms, as well as for evaluating various scalability schemes. It is possible to model application behavior exactly according to the specifications since applications are usually implemented in software in real world. The more challenging aspect of node behavior modeling is that of the communi- cation layers (transmitter and receiver) below application. Nevertheless, transmitter and receiver behavior can generally be modeled with a higher fidelity than the channel model. The reason is that most of the communication components can either be exactly implemented based on the protocols, or abstracted using data derived from tests. This is in particular true for higher layers of the protocol stack. Generally, the behavior of MAC layer and above is implemented exactly; however, the physical layer behavior may need to be abstracted. In the case of DSRC, the receiver and sender behavior at the lowest parts of the physical layer has to be abstracted to frame level behavior, to allow matching it to the channel model behavior. If channel model variations at bit or symbol level is used, the same has to be considered for the node model. This is however, not a common approach, although some works exist [21]. With frame level modeling of the node behavior, the focus is on specific actions that are taken by the physical layer at the boundaries of a frame. Receiver decoding probabilities, interference and noise effects are also calculated at the frame boundary and at the end a probability for successfully decoding a frame is calculated. The frame level modeling can be improved by considering specific receiver and decoding events that happen inside a frame. We call this “sub-frame modeling” of a receiver behavior. With subframe modeling, different parts of a frame that change the processing steps and decisions inside a receiver will be considered. This is in particular important for DSRC transceivers, since there are a few decisions that are made based on successful decoding of portions of a frame. Subframe modeling can significantly improve the modeling accuracy, while having only a small computational overhead. In later sections of the chapter, details of the sub frame modeling will be discussed.
Efficient and High Fidelity DSRC Simulation 221 Modeling the node behavior at the physical layer is mostly focused on modeling receiver behavior, since PHY transmitter behavior is generally assumed determinis- tic and is not impacted by the variations that result from the channel behavior. As a result, the transmitter model is in general much simpler and is modeled as simple gain values for DSRC in most simulators. The MAC behavior is, however, different and full implementation of the rather complex 802.11 MAC for DSRC is required for both sender and receiver behaviors. Mobility and Environment Models An important factor that determines the channel behavior, and consequently the performance of a receiver is the physical property of the environment between a pair of sender and receiver. Distance between nodes, presence or lack of line of sight (LOS), and possibility of occluding objects are some pf the properties that can be determined from position of nodes and blocking objects. For DSRC, the nodes are intrinsically mobile and the channel between each pair of vehicles will change (slowly in comparison to communication channel parameters) as vehicles move. The movement of vehicles is usually imported from other sources. Vehicle trajectories can be imported from traffic simulators (e.g., SUMO or VISSIM), or from trajectory logs of actual vehicle movement. This aspect of the simulation is generally handled separately from the communication part and will not be further elaborated in this chapter. Node Model As it was mentioned, most of the complexity in modeling the DSRC node behavior is in modeling the receiver behavior and in particular in relation to abstract frame or sub frame level processing. The MAC behavior is generally implemented exactly, although it is impacted by how the PHY events are abstracted. This is discussed in detail model in this section. The PHY transmitter behavior for DSRC is simply modeled as power gains; MAC behavior of the DSRC transmitter is also implemented exactly as it is described by the 802.11 protocol. To understand how the PHY receiver is modeled, we note that a DSRC frame is comprised of several parts. Namely, the preamble, PLCP header and payload. Further details on MAC header and parts are not used in sub frame processing. The specific actions of the receiver are taken at the boundaries of these sections of frames. As a result, the processing of a packet can be done in a discrete event simulator, like ns-3, in an efficient way and using event scheduling at the moments that each section of the frame is expected to be available at a receiver. Each receiver in the wireless environment independently treats the incoming frame at the moments that these sections of a frame is received. The event times in ns-3 are calculated by
222 Y. P. Fallah and S. M. Osman Gani considering the transmission time and propagation delay (computed using distance between sender and receiver). The specific actions upon reception of each section of a frame will depend on whether the received portion is usable or decodable by the receiver. To determine the success in reception, and in the absence of bit level processing, the receiver will have to use quantities such as received signal strength (RSS) and signal to noise and interference ration (SINR). In most models, the value of SINR is used, along with some error or reception probability model, to determine if the received section of the frame is decodable. In the next subsections we will elaborate on how SINR is calculated and how frame decoding is handled. It must be noted that this description is based on our DSRC simulator which was built by enhancing and correcting existing ns-3 models. Therefore, most of the explanations are dedicated to how existing models need to be enhanced to properly support DSRC. Frame Structure As per the IEEE 802.11 standard [15], the PLCP Protocol Data Unit (PPDU) in OFDM systems has three main parts (Fig. 2): • PLCP Preamble: This field is used at the receiver end to synchronize the demodulator. The PLCP preamble consists of 12 symbols: 10 short symbols and 2 long symbols to mark the start of a frame. For 10 MHz channel spacing, the duration of the preamble is 32 μs. • SIGNAL: The preamble is followed by SIGNAL and DATA. SIGNAL is one symbol long and contains LENGTH and RATE information of the frame, which is part of the PLCP header. Fig. 2 PPDU format [15]
Efficient and High Fidelity DSRC Simulation 223 Fig. 3 Different parts of an OFDM frame along with associated action events • DATA: The DATA part contains the Service field of the PLCP header, and the rest of the frame with pad and tail bits. Evidently, there are three main parts of an OFDM frame based on PPDU format shown in Fig. 2. Therefore, we emulate an OFDM frame by dividing the total frame duration into three parts: the PLCP preamble, PLCP header and the frame payload. Corresponding processing events are scheduled at the boundaries of each part. Here we loosely refer to SIGNAL decoding as PLCP header decoding since this is the part that significantly impacts how the rest of the decoding happens in our model. Figure 3 illustrates the three parts along with the associated action events [22]. Receiver Frame Processing Model As illustrated in Fig. 4, the frame processing model is required to maintain an interference list, as well as a state machine that defines the instantaneous state of the receiver. The high-level reception flow diagram shows how an incoming frame is treated based on the current PHY state. In the following subsections, we detail the interference model, the PHY state determination, and frame detection and decoding steps. Interference Model To accommodate physical carrier sensing, energy detection and to model impact of interference on signal decoding, receiver model keeps track of cumulative interference by maintaining a list of all signals that are currently on the medium. This list is populated by adding the incoming signal which can be described by the received power RxPower, and the start and end time of a signal that has just arrived at the receiver, assuming that the RxPower does not vary over the frame lifetime. The Interference list helps to calculate instantaneous cumulative interference by simply summing all signals that are active at that time. Also, calculation of signal- to-interference-and-noise ratio (SINR) is possible for individual signals, which is frequently used for checking reception probability and making decision about reception continuation or frame capture. Determination of SINR for a signal of interest is required for various frame detection and decoding steps. For example, the arrival of a new signal during the
224 Y. P. Fallah and S. M. Osman Gani Start Signal Reception Add Signal to Interference List Check Receiver PHY State IDLE or CCA_BUSY RX TX Start Frame Frame Capture Drop Incoming Reception Frame Fig. 4 Receiver frame process model (big picture) reception of another signal requires a SINR check to decide whether the newer frame should be captured or not. To accommodate the calculation of SINR for any signal, a subroutine is needed which returns the interference of the signals that overlap with a signal or a series of signals. During SINR calculation, noise floor and noise figure are also considered. Noise floor is a measure of all unwanted background noise present in the system. In general, thermal noise is modeled as Nt = KTB, where K is Boltzmann constant, T is temperature in Kelvin, and B is the signal bandwidth in Hz. Noise figure, NF, is a measure of degradation of SNR caused by the components in the radio signal chain, for a given bandwidth. Therefore, the total noise floor is calculated by noise f loor = N F ∗ Nt SINR of signal i is then calculated by SI N Ri = noise floor + RxP oweri j =signal overlaps with i RxP owerj Where RxPoweri and RxPowerj are in Watts. An example SINR computations for a signal of interest (RED signal) at different instance of time is illustrated in Fig. 5.
Efficient and High Fidelity DSRC Simulation 225 Fig. 5 An example SINR computations for a signal of interest (RED signal) at different instances of time PHY State The IEEE 802.11 PHY can be modeled based on YANS [14], as it is done in the popular ns-3 simulator. In this model, the state machine of the Wifi transceiver has the following four states: • Transmission mode (TX): PHY is currently transmitting a signal. The TX state cannot be interrupted by any other events • Reception mode (RX): PHY is synchronized to a signal and currently receiving that signal • Idle mode (IDLE): There is no signal sensed in the channel and the PHY is idle • Clear channel assessment busy mode (CCA_BUSY): The physical layer is not in the TX or RX state, but the total power in the medium is higher than the (non- OFDM) CCA Energy Threshold PHY states (IDLE or CCA_BUSY) are determined using Clear channel assess- ment (CCA) function, which is defined in the IEEE 802.11 standard. CCA involves two related functionalities: Carrier sensing and Energy Detection. Carrier sensing refers to receiver’s ability to detect and decode an OFDM signal on the channel. Energy detection refers to the receiver ability to detect non-OFDM energy on the channel. Based on hardware tests that were done as part of a project by Crash Avoidance Metrics Partnership (CAMP) [30, 22], the start of a valid OFDM frame requires RSS of approximately −94 dBm or greater. The CS/CCA should indicate busy medium when preamble of a valid OFDM frame starts with a RSS of −94 dBm or greater, implying that the receiver has synced to an OFDM frame. If the preamble and PLCP header are successfully received (we will later discuss how to determine this), the channel should be held busy for the duration of the frame, which can be calculated based on LENGTH and RATE information of the PLCP header. If the preamble is not successfully receiver, the CCA function falls back to energy level detection. In this case, the receiver shall hold CCA signal busy for the duration
226 Y. P. Fallah and S. M. Osman Gani for which the total energy on the channel is greater than a specific threshold. From the hardware test mentioned above, this threshold is found to be around −82 dBm. It must be noted that the values of −94 dBm and −82 dBm are found for specific hardwares (Denso WAVE Safety Unit and Savari On-Board Units) and these particular thresholds may change for other or future devices. Nevertheless, the values are expected to be more or less similar for all 802.11 devices since the values are in fact related to the underlying 802.11 standard designs. Frame Detection and Decoding Steps Frame detection and decoding starts with detection of preamble. Following a successful preamble detection and synchronization, the physical layer synchronizes to the incoming frame and goes to RX state. The preamble detection step is implemented by scheduling an event 32 μs after start of a frame. This event compares the observed SINR with the preamble successful decoding threshold (PSDT). If the SINR is less than the PSDT, receiver could not successfully decode the preamble. Therefore, the current frame that is being received is dropped and the PHY state is decided based on cumulative signal strength. If the total signal strength is below the CCA energy detection threshold, PHY goes to IDLE. Otherwise, PHY goes to CCA_BUSY. If the measured SINR is high enough for the device to be able to detect the preamble, the receiver enters the next phase, which is the PLCP header decoding. PLCP header decoding event is required to ensure that the receiver has correctly decoded the PLCP header. This check is scheduled at the end of the PLCP header of the frame. The observed SINR is compared with the PLCP header decoding threshold. If SINR is above the threshold, the header is considered successfully decoded. If header decoding fails, the ongoing reception is aborted, and cumulative signal strength is checked to decide about the PHY state transition. Like before, if the total signal strength is below the CCA energy detection threshold, PHY goes to IDLE and it goes to CCA_BUSY, otherwise. Figures 6 and 7 illustrate the scheduling of preamble detection and PLCP header check when a new frame arrives, and the receiver state is IDLE or CCA_BUSY. ab RxPower > OFDM No Signal Detection Yes Total Power > CCA Energy Threshold Threshold No Yes Energy Detect PREAMBLE CHECK Event PHY State = IDLE PHY State = CCA_BUSY Fig. 6 Scheduling of PLCP preamble check event, and energy detect flowchart. (a) Scheduling preamble check event. (b) Energy detection
Efficient and High Fidelity DSRC Simulation 227 ab SINR > Preamble No SINR > PLCP Header No Successful Decoding Drop Receiving Decoding Threshold Drop Receiving Threshold Packet Yes Packet Yes Energy Detect Frame Decoding Energy Detect Event PLCP Header CHECK Event Fig. 7 Preamble and PLCP header check events. (a) Preamble check event. (b) PLCP header check event Frame decoding event is the last step that is scheduled following successful decoding of PLCP header. After successful decoding of PLCP header, the receiver stays in RX mode and will schedule another event at the end of the frame to determine whether the payload parts can be successfully decoded. It is however possible that before the frame end event happens, other events occur due to interference, as described in the next subsection. Successful decoding of a frame, at the frame end event, will depend on the amount of interference that occur during the reception of the frame. The probability of successful decoding is determined by calculating SINR values for each chunk of the frame that suffers interference, and calculating the probability of reception based on a predetermined “error model”. The error model will specify the probability of reception as a function of SINR. The function is derived using bit error rate (BER) vs. SINR relationships for different modulation and coding schemes [18]. Either theoretical formulation, or hardware tests may be used to determine the error model. Our work in [22] used the hardware test mentioned earlier, to determine the error model. Threshold Value Selection PSDT and PLCP header decoding threshold are used to decide if the frame reception should continue at the end of the preamble and PLCP header, respectively. PSDT of 3 dB is chosen based on the frame success rate of a frame which is exactly equal to the preamble duration of an OFDM signal in length that is transmitted at 6Mbps. We observe that the transition to successful frame reception occurs when the SINR is above 3 dB. The same procedure is used for PLCP header decoding threshold and its value is chosen to be 2 dB. These values have been then fine-tuned based on experiments with field data and comparison of simulation results (through trial and error) [22]. It must be noted that it is possible to use the probabilistic approach of using an error model and SINR value instead of the above specific
228 Y. P. Fallah and S. M. Osman Gani thresholds. Our observations have shown that the results are very similar and the added complexity of probabilistic check for preamble or PLCP reception may not be necessary. The reason for similar results is that theoretical error model formulations have a sharp transition in probability of success around threshold values and the effect of using the probability model instead of a specific threshold is minimal. Frame Decoding: Error Model One of the factors that impacts the probability of decoding a frame, in addition to SINR and decoder specifics, is the length of a data frame. In fact, the BER vs. SINR relationship only depends on SINR and receiver parameters; but PER (packet error rate) is a function of BER and the length of the frame (loosely called packet here). While the general approach is to derive the PER using theoretical approaches based on modulation and coding choices, the results are often somewhat different from what actual devices achieve. We observed this in some of our previous works [22]. The error models (PER vs. SINR) that we obtained from hardware tests mentioned earlier are depicted in Fig. 8. The model was obtained for a single frame size (BSM size) with DSRC 6 Mbps modulation and coding option. The model can be used to approximate error model for different frame sizes as depicted. Using curve fitting, a general format of the model is obtained as following: a × erf x − b + d c Where a = 0.4997, b = 3.557, c = 1.292, and d = 0.5. Fig. 8 Empirical error model for different-sized frames
Efficient and High Fidelity DSRC Simulation 229 Handling Partial Interference An issue that arises when an empirical model as in Fig. 8 is used, is the issue of partial interference. When only parts of two frames overlap, the SINR values will be different for different chunks of a frame. The theoretical error model that uses BER to calculate PER for any frame size can handle this issue by separately determining the probability of reception. However, this may be somewhat problematic as the error correction schemes are usually applied to the entire frame and not to chunks. Another issue is that the theoretical models are not very accurate. If empirical models are used, the value of PER has to be recalculated for considerably smaller chunks and the method of obtaining BER from PER and then obtaining PER for a different frame (chunk) size is also not very accurate. As a result, we may resort to a simpler method of determining frame reception by considering the worst-case scenario and finding the minimum SINR of the received frame during its reception. This conservative approach ensures that if a BSM can be received with the minimum SINR it experiences during the reception process, all other experienced SINR values would also ensure a successful reception. Frame Capture Feature An important feature of 802.11 receivers is the possibility of frame capture. Frame capture allows a wireless receiver to lock on to a stronger signal in the presence of other signals (interferences) regardless of its arrival time. It occurs when two or more signals overlap with each other. In the simplest form of a receiver model, if multiple signals interfere with each other, the receiver cannot decode any of the signals because they are garbled. But in real-world wireless devices, receiver can decode the stronger signal provided that it is strong enough for successful decoding. In DSRC based safety communications, frame capture happens a lot because the scenarios (hidden node collision) that can lead to overlapping of multiple signals are inherent in vehicular networks. DSRC enabled vehicles exchange information by periodically broadcasting BSMs. Since broadcast does not use RTS/CTS mechanism for node coordination, nodes are less aware of other ongoing transmissions in their surrounding areas. And thus, hidden terminal problems become common in vehicular networks. Due to the presence of hidden terminals, multiple frames can arrive at a receiver almost simultaneously and can lead the receiver to capture one of them. Though hidden nodes are responsible for most of the capture scenarios, there is another scenario, where two signals can overlap. If the backoff counter of two stations reach zero at the same time, they can start transmission simultaneously. Frame Capture Scenarios In previous subsection we have explained the capture effect and the scenarios when it occurs. Now we classify those scenarios, and for classification purpose, we assume
230 Y. P. Fallah and S. M. Osman Gani a Δt Preamble Second Frame Preamble First Frame time b First Frame Preamble Δt Preamble Second Frame time c First Frame Preamble Δt Preamble Second Frame time Fig. 9 Frame capture scenarios. (a) Sender first capture (SFC). (b) Sender last capture (SLC): preamble capture. (c) Sender last capture (SLC): frame body or payload capture that only two frames are in collision, and the second frame arrives during the first frame’s reception. Considering the arrival timing and signal strength, physical layer frame capture can be classified (see Fig. 9): 1. Sender First Capture (SFC) Stronger frame’s preamble detection is successful, but the payload suffers because of interference. The arrival timing of the weaker signal is not important here as it arrives after the receiver is locked to the stronger one. 2. Sender Last Capture (SLC) In this capture, first frame (weaker frame) is received till the arrival of the stronger frame. When the second frame arrives, the first one is garbled; the receiver ceases receiving that frame and locked on to the second frame. Based on arrival time, SLC can be further classified in two cases; preamble capture and frame body/payload capture. Frame Capture Implementation Depending on the arrival time of the frame and the current PHY state, receiver uses different thresholds to capture an incoming frame.
Efficient and High Fidelity DSRC Simulation 231 • Preamble capture: If the new signal arrives during the reception of the preamble of another frame, SINR of the incoming signal is checked against the preamble successful capture threshold (PSCT). The receiver drops the receiving frame and synchronizes to the newly arrived frame if SINR is above PSCT. SINR below PSCT suggests that currently receiving signal is strong enough to be decoded and the receiver continues the reception. Figure 10 illustrates the preamble capture scenario. • Data/payload capture: If the arrival of the new signal occurs during the payload of a receiving frame, SINR is checked against the data capture threshold (DCT). The receiver drops the currently receiving frame if SINR is above DCT. SINR below DCT suggests that the incoming signal strength is not strong enough for the receiver to lock to it, and the newly arrived frame should be discarded. Figure 11 shows the data capture scenario. Figure 12 illustrates the frame capture along with associated action events. Preamble Duration CCA Frame A (in RX) OSDT Noise CCA Frame B arrives OSDT within A’s Preamble CCA Total Power OSDT PSCT SINR for Frame B Sync to Frame B: (SINR(B) > Preamble Successful Capture Threshold) Fig. 10 Preamble capture
232 Y. P. Fallah and S. M. Osman Gani Preamble Duration Frame A (in RX) Noise CCA Frame B (arrives OSDT after A’s preamble) CCA Total Power OSDT CCA OSDT DCT SINR for Frame B Sync to Frame B: (SINR(B) > Data Capture Threshold) Fig. 11 Data/payload capture Yes RX Duration < No Preamble Duration SINR > Preamble No Successful Capture Yes SINR > Data Capture Yes Threshold No Threshold Drop Currently Continue Receiving Drop Currently Continue Receiving Receiving Frame Current Frame Receiving Frame Current Frame Switch to New Switch to New Frame Frame PREAMBLE PREAMBLE CHECK Event CHECK Event Fig. 12 Frame capture flowchart Threshold Value Selection For capture effect implementation we use two threshold values to make decision about frame switching.
Efficient and High Fidelity DSRC Simulation 233 • A value of 8 dB is derived from the V2V-Interoperability Project radio hardware testing for DCT. A similar value of 7–8 dB of SINR for 100% frame success rate has been reported about validation of OFDM error rate model [18]. • Simulations for different PSCTs have been run to get the preamble successful capture threshold (PSCT) while the other threshold values were kept fixed. The results are compared with the V2V-Interoperability Project field tests to find a match. Using a value of 7 dB for PSCT we find an acceptable match with the field result. Channel Model Considering the natural phenomena that can deteriorate signals, three elements are usually considered in describing a wireless channel: large-scale path loss, shadowing or large-scale fading, and small-scale fading (Fig. 13). Large-scale path loss describes the deterministic signal attenuation at a specific distance. Shadowing, in vehicular communications, occurs when signals have to pass through large objects obstructing the sender and the receiver. Vehicular networks are very dynamic in nature, and thus, shadowing changes quickly over time. Small-scale fading, sometimes just referred to as fading, captures the signal strength changes due to Fig. 13 Effects of channel components
234 Y. P. Fallah and S. M. Osman Gani vehicle movements (e.g., effects of multipath, Doppler shift, etc.). All these models work together and define a wireless channel. Usually it is very difficult to theoretically derive an exact channel model for vehicular environments due to the physical complexity and fast changing dynamics of the environment [20]. As a result, models are derived from empirical data and measurements done in different environments [25–27]. There are many approaches to derive a model from empirical data collected from fields. In simpler forms, received signal strength indicators (RSSIs) of the received frames can be used to derive a model for how received power changes as a function of distance between sender and receiver. Since large scale and small scale fading both model the variation in signal levels, they are sometimes combined if the underlying empirical data does not have enough data to separately derive each [27]. Here we briefly look at such models. The subject of channel modeling is outside the scope of this chapter; neverthe- less, channel model components are important elements of a simulator and their use and implementation will be discussed in the next section. Channel Model Components The channel model in frame level simulation is a mathematical relationship that describes how received signal power of a frame changes as a function of the environment features. In simplest form, the distance between sender and receiver is assumed as the most important factor and is used as the only parameter describing the environment. The channel model is then reduced to a formula expressing propagation loss (power loss) as a function of distance. For example, the received power from a sender at an arbitrary distance d is expressed as Pr (d) = Pt − LLS(d) + gP ,dB (dB) , (1) where Pt is the transmission power in dB, LLS(d) is the deterministic, large-scale path loss at distance d, and gP, dB is the random, small-scale fading in dB. Figure 14 shows all RSSI measurements versus distance d for all vehicles in the field test scenario in a relatively low traffic highway [27]. For this test the rate of transmission power was set Pt = 20 dBm. Each of the components in Eq. (1) can be approximated using the data collected and shown in Fig. 14. Large-Scale Path-Loss Model The first component of the channel loss model is the large-scale loss. In most vehicular environments, where the signal is not completely blocked by traffic, a two-ray propagation model [19] exist and is used to model the deterministic, large- scale path loss component. This model is recently validated for vehicular networks
Efficient and High Fidelity DSRC Simulation 235 0 -10 All RSSI Mean + STD -20 Mean Mean - STD -30 RSSI (dBm) -40 -50 -60 -70 -80 -90 -100 200 400 600 800 1000 1200 1400 0 Distance (meters) Fig. 14 Sample RSSI measurements versus distance for an example of a low traffic highway (all vehicles transmitting with a power of 20 dBm) Fig. 15 Conceptual ht dlos framework of the two-ray interference model for d large-scale path loss [19] θi dref hr and takes into account the interference caused by a single ground reflection ray at the receiver, as shown in Fig. 15. Based on the two-ray interference model, the distance-dependent, large-scale path-loss in the wireless channel of a vehicular network can be found as [19] LLS (d; α, r ) = 10 α log d 1 + Λeiφ −1 (dB) , (2) 4π λ where α denotes the path-loss exponent, and λ = c is the signal wavelength f corresponding to the transmitted signal with center frequency f that is propagating in the environment with the speed of c. In the above equation, the reflection coefficient Λ can be found as
236 Y. P. Fallah and S. M. Osman Gani Λ = sin θ − r − cos2θ , (3) sin θ + r − cos2θ where r is a fixed, unit-less constant dependent on the reflection medium, sin θ = ht +hr and cos θ = d , dref = d2 + (ht + hr )2 shown in Fig. 15, and ht and dref dref hr are the heights of the transmitter antenna and receiver antenna, respectively. Furthermore, the phase difference of the two interfering rays φ can be found as φ = 2π dlos − dref , (4) λ where dlos = d2 + (ht − hr )2 is shown in Fig. 15. Two-ray ground reflection path-loss model has two unknown parameters that could be found based on the collected empirical data: path-loss exponent, α and r, which is a fixed, unit-less constant dependent on the reflection medium. An interested reader is referred to [25–27] for more information on how these parameters are derived from empirical data. Fading Model Previous studies have found that small-scale fading of wireless channel mainly have a Nakagami-m distribution [25]. This distribution has the ability to model a wide range of small-scale fading scenarios from strong line-of-sight (LoS) and Rician- distributed fading (larger values of m > 1), to non-LoS and Rayleigh-distributed fading (unit value for parameter m). For large distances (e.g., beyond the Fresnel distance), Weibull distribution is used as it is widely used in the literature [28]. As it was mentioned earlier, shadowing and small-scale fading effects are sometimes combined and modeled as one factor, in particular when the empirical data does not provide enough details to model them separately. Overall, fading model can be viewed as a way of describing signal strength variations and randomness. Channel Model Chaining After modeling all the channel components, they can be chained together to get the final RxPower as shown in Fig. 16. Here, the received signal strength is calculated by passing the transmit power through a chain of channel model components and other deterministic factors (such as antenna gain) not shown here.
Efficient and High Fidelity DSRC Simulation 237 TxPower Deterministic Shadowing Fading Model RxPower Path Loss Model Model Fig. 16 Propagation model chaining Receiver Frame Process Model Validation Verifying the model developed for a receiver is less straightforward than that of the propagation. Propagation model can be verified by comparing RSS values generated by the model, with that of the field tests. To evaluate receiver model, we need performance metrics that can be measured both in the field and in simulation. Two of the possible metrics are: Channel Busy Ratio (CBR) is the fraction of time the communication channel is sensed busy by a node during a predefined period of time. In the field experiments, CBR was calculated every 100 ms. Packet Error Ratio (PER) is the ratio of missed data packets to the total number of packets transmitted during a predefined time window. This metric is calculated for each sender-receiver pair. Since these metrics are relatively high level, meaning that they are measured above the PHY layer, all of the internal mechanisms of the PHY, as well as the propagation model will impact them. In particular, any inaccuracy in propagation channel modeling will be directly inserted into the measurements of PER and CBR. As a result, these metrics should only be used if a good and acceptable channel model is available. Taking this into consideration, these metrics will provide a way of validating the combined effect of all of the internal PHY models and mechanisms (from error models, sub frame processing, thresholds, etc.). To remove as much uncertainty as possible, some of the internal mechanisms, thresholds and parameters are derived from hardware test and the other ones are adjusted by trial and error (refer to the node model for details). In this section, we discuss the validation and verification results. Verification of Frame Capture Implementation To verify the correctness of our frame capture implementation, we set up a three- node simulation scenario (Fig. 17) as follows: • Node A (moving node) moves towards node 1 (receiver node) at a speed of 4 m/s. Node A works both as receiver and sender. • Node B remains stationary and receives packet form A and C • Node C remains stationary and receives and sends out packet
238 Y. P. Fallah and S. M. Osman Gani fixed Sender Receiver Sender A 1000 m B 2000 m C varying Fig. 17 Three-node scenario for capture effect verification • Both senders are initially placed in such way that they are hidden from each other. Node A always starts transmission earlier than Node B. As Node A gets closer to the receiver B, its signals get stronger. In this simulation setup we use FriisPropagationLossModel, and no fading models, for the sake of analysis simplicity. As discussed earlier we can write the Friis equation as follows Pt = Gt Gr λ2 , (5) Pr (4π d)2L where Pr and Pt are the received and transmitted power in watt, respectively; Gr and Gt are the receiver and transmitter antenna gain, L is a dimensionless system loss coefficient, λ is the wavelength, and d is the receiver-transmitter separation distance. From Eq. (5) d can be solved as d = λ Pt Gt Gr (6) 4π Pr L λ is determined as c , where c = speed of light = 3 × 108 m/s and f is the channel f frequency band. For DSRC channel, λ is calculated as λ = 3∗108 m/s = 0.0509 m. 5.89 GH z In this test, default transmission power, Pt is set to 20 dBm and Energy detection threshold is set to −94 dBm. Thus, using Pr = −94 dBm in Eq. (2) will give us the range, d of a transmitter in this simulation setup; calculated d is 2032 m. Node A is hidden from Node C till they are 2032 meters apart.
Efficient and High Fidelity DSRC Simulation 239 Calculated PER at Receiver B for Sender A 1 0.8 0.6 PER 0.4 0.2 0 0 1000 900 800 700 600 500 400 300 200 100 A-B Separation Distance (meters) Fig. 18 PER vs. separation distance of sender and interferer Depending on the timing relation of the two transmitted frames from sender A and sender C, and receiver B observes preamble or payload capture effect. Node C starts transmitting first; the signal reaches B after 6.6 μs. Node A starts sending after 6.6 μs to make sure its signal reaches receiver B after C’s signal, and thus triggers capture effect. The A-C separation distance decreases as node A moves towards C. When A-C separation distance becomes 2032 m, A can hear signals from C, and CSMA/CA mechanism does not allow A’s transmission. Below this separation distance threshold, both signals could be received correctly. As shown in Fig. 18, PER of sender A decreases as it gets closer to receiver B. Overall Receiver Model Validation To validate the overall DSRC receiver model, we simulate the same VANET scenario that was used by V2V-Interoperability Project to collect data. The mobility traces are extracted for the vehicles from the field test GPS data. Ten percent of the vehicles are loggers and distributed uniformly among other vehicles. They are capable of recording channel busy ratio (CBR), GPS data, and various transmissions (TX) and reception (RX) logs. Table 1 summarizes our simulation settings. We performed ns-3 simulations for two different configurations: frame-level ns- 3 without frame capture, and enhanced subframe-level ns-3, i.e., ns-3 with capture effect, preamble and header decoding features. To compare the results from frame-level and enhanced subframe-level ns-3 physical layer implementation, we look into two metrics: Channel Busy Ratio (CBR) and Packet Error Ratio (PER). CBR is calculated every 100 ms. To calculate PER,
240 Y. P. Fallah and S. M. Osman Gani Table 1 Simulation settings Parameter Value in ns-3 Simulation run time Receiver noise figure 250 s OFDM signal detection threshold (OSDT) 6 dB CCA energy threshold −94 dBm Preamble successful decoding threshold (PSDT) −82 dBm PLCP header decoding threshold 3.0 dB Preamble successful capture threshold (PSCT) 2.0 dB Data capture threshold (DCT) 7.0 dB 8.0 dB Fig. 19 Frame-level vs. sub-frame level simulation comparison we use RX logs. We employ a sliding window approach with a 2 s window and 1 s sub-window. PER values for logger nodes are calculated w.r.t. node 1 (a moving node). Here we use 40-m distance bins. Figure 19 show the PER values for all vehicles that are within a 1000-m range of vehicle 1. These figures indicate that messages from closer vehicles are received with high probability. As the distance between the sender and receiver increases, the PER is expected to go up. PER curves become more and more similar to the field PERs as the subframe-level action events and frame capture features are added one at a time. CBR is another metric that is measured at each logging vehicle. Each vehicle records the fraction of time it sensed the channel busy over a 100 ms period.
Efficient and High Fidelity DSRC Simulation 241 100 100 90 80 80 70 CBR(%) 60 60 50 40 40 30 20 20 1.3843 10 0 1.3843 30 x 1012 1.3843 2500 2000 Time stamps 1.3843 1000 1500 X-Position Fig. 20 CBR for all moving nodes (field) Fig. 21 CBR at node 1: (left) Frame-level simulation, (center) Field test, (right) Subframe-level simulation Figure 20 shows the field CBR for all moving nodes. Figure 21 offers a clearer insight into the CBR curve. Here, CBR of node 1 from the field test, frame-level ns-3 and enhanced subframe-level ns-3 are plotted and compared with each other. CBR measured at vehicle 1 for both the frame-level ns-3 and enhanced subframe- level ns-3 are slightly higher than the field result. However, CBRs from enhanced subframe-level ns-3 show a better match with the field test. Conclusion This chapter presented an overview of DSRC simulation process, describing specific components of simulation which include node and propagation modeling. Node behavior modeling has been presented with a focus on the most commonly used approach of frame or sub frame based simulation. We described methods used by popular discrete event simulators such as ns-3, and presented several corrections and enhancements that were made to achieve higher fidelity of simulation for DSRC.
242 Y. P. Fallah and S. M. Osman Gani Channel modeling approaches were also briefly discussed, and the steps of frame based channel model implementation has been presented in detail. To describe the channel propagation model, a specific model developed from large scale field test data has been explained. The overall simulation model for the channel and node behavior was validated and shown to be more accurate than the frame-level simulator. The model is used in several studies and has also been contributed to the ns-3 community for its on-going effort of developing the simulator. References 1. R. Sengupta, S. Rezaei, S. E. Shladover, J. A. Misener, S. Dickey, H. Krishnan, “Cooperative collision warning systems: concept definition and experimental implementation,” Journal of Intelligent Transportation Systems, vol. 11, No. 3, pp. 143–155, 2007. 2. K. Dar, M. Bakhouya, J. Gaber, M. Wack and P. Lorenz, “Wireless communication tech- nologies for ITS applications [topics in automotive networking],” IEEE Communications Magazine, 48(5), pp. 156–162, 2010. 3. Bai F., Krishnan H., “Reliability Analysis of DSRC Wireless Communication for Vehicle Safety Applications,” 2006 IEEE Intelligent Transportation Systems Conference, Toronto, Ont., 2006, pp. 355–362. 4. Tahmasbi-Sarvestani A., Fallah Y.P., Kulathumani V., “Network-Aware Double-Layer Distance-Dependent Broadcast Protocol for VANETs,” in Vehicular Technology, IEEE Trans- actions on, vol. 64, no. 12, pp. 5536–5546, Dec. 2015 5. Q. Chen, D. Jiang, V. Taliwal, and L. Delgrossi, “IEEE 802.11 based vehicular communication simulation design for NS2,” in Mobile Computing and Networking, 2006, pp. 50–56. 6. Q. Chen, F. Schmidt-Eisenlohr, D. Jiang, M. Torrent-Moreno, L. Delgrossi, and H. Hartenstein, “Overhaul of Ieee 802.11 Modeling and Simulation in Ns-2,” in Proceedings of the 10th ACM Symposium on Modeling, Analysis, and Simulation of Wireless and Mobile Systems, New York, NY, USA, 2007, pp. 159–168. 7. J. Lee, W. Kim, S.-J. Lee, D. Jo, J. Ryu, T. Kwon, and Y. Choi, “An Experimental Study on the Capture Effect in 802.11a Networks,” in Proceedings of the Second ACM International Workshop on Wireless Network Testbeds, Experimental Evaluation and Characterization, New York, NY, USA, 2007, pp. 19–26. 8. J. Ryu, J. Lee, S.-J. Lee, and T. Kwon, “Revamping the IEEE 802.11a PHY Simulation Models,” in Proceedings of the 11th International Symposium on Modeling, Analysis and Simulation of Wireless and Mobile Systems, New York, NY, USA, 2008, pp. 28–36. 9. J. Lee, J. Ryu, S.-J. Lee, and T. T. Kwon, “Improved modeling of IEEE 802.11a PHY through fine-grained measurements,” Comput. Netw., vol. 54, no. 4, pp. 641–657, 2010. 10. “QualNet Network Simulator,” http://www.scalable-networks.com/. 11. S. Papanastasiou, J. Mittag, E. Strom, and H. Hartenstein, “Bridging the gap between physical layer emulation and network simulation,” in Proceedings of the 2010 WCNC, April 2010. 12. J. Mittag, S. Papanatasiou, H. Hartenstein, and E. Strom, “Enabling accurate cross- layerPHY/MAC/NET simulation studies of vehicular communication networks,” Proc. IEEE Special Issue on Vehicular Communications, June 2011. 13. “ns-3 Network Simulator,” http://www.nsnam.org/. 14. M. Lacage and T. R. Henderson, “Yet Another Network Simulator,” in Proceeding from the 2006 Workshop on NS-2: the IP Network Simulator. Pisa, Italy: ACM, 2006, p. 12.
Efficient and High Fidelity DSRC Simulation 243 15. “IEEE Standard for Information technology–Telecommunications and information exchange between systems Local and metropolitan area networks–Specific requirements Part 11: Wire- less LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications,” IEEE Std 80211-2012 Revis. IEEE Std 80211-2007, pp. 1–2793, Mar. 2012. 16. Moradi-Pari, E.; Tahmasbi-Sarvestani, A.; Fallah, Y.P., “A hybrid systems approach to modeling real-Time situation-awareness component of networked crash avoidance systems,” in Systems Journal, IEEE, vol. PP, no. 99, pp. 1–10. 17. Moradi-Pari, E.; Tahmasbi-Sarvestani, A.; Fallah, Y.P., “Modeling communication and esti- mation processes of automated crash avoidance systems,” 2013 IEEE International Systems Conference (SysCon), Orlando, FL, 2013, pp. 681–687. 18. Guangyu Pei and Thomas R. Henderson, “Validation of OFDM error rate model in ns-3,” Boeing Research Technology, pp. 1–15, 2010. 19. C. Sommer, S. Joerer, F. Dressler, “On the applicability of two-ray path loss models for vehicular network simulation,” IEEE Vehicular Networking Conference (VNC), 2012. 20. L. Cheng, B.E. Henty, D.D. Stancil, F. Bai, and P. Mudalige, “Mobile vehicle-to-vehicle narrow-band channel measurement and characterization of the 5.9 GHz dedicated short range communication (DSRC) frequency band,” IEEE Journal on Selected Areas in Communica- tions, vol. 25, no. 8, pp. 1501–1516, October 2007. 21. J. Lu, G. Han, J. Wang, B. Li and W. Dou, “HCEM: A new class of bit-level hybrid channel error model,” 2012 3rd IEEE International Conference on Network Infrastructure and Digital Content, Beijing, 2012, pp. 8–12. 22. S. M. O. Gani, A. Tahmasbi-Sarvestani, M. Fanaei and Y. P. Fallah, “High fidelity DSRC receiver model for ns-3 simulation using large-scale field data,” 2016 IEEE Wireless Commu- nications and Networking Conference, Doha, 2016, pp. 1–6. 23. SAE International, “Surface Vehicle Standard – On-Board System Requirements for V2V Safety Communications,” J2945TM/1, Issued 2016-03 24. “Dedicated Short Range Communications (DSRC) Message Set Dictionary, Std. J2735”, Mar. 2016. 25. L. Cheng, B. E. Henty, D. D. Stancil, F. Bai and P. Mudalige, “Mobile Vehicle-to-Vehicle Narrow-Band Channel Measurement and Characterization of the 5.9 GHz Dedicated Short Range Communication (DSRC) Frequency Band,” in IEEE Journal on Selected Areas in Communications, vol. 25, no. 8, pp. 1501–1516, Oct. 2007. 26. M. Boban, T. T. V. Vinhoza, M. Ferreira, J. Barros and O. K. Tonguz, “Impact of Vehicles as Obstacles in Vehicular Ad Hoc Networks,” in IEEE Journal on Selected Areas in Communications, vol. 29, no. 1, pp. 15–28, January 2011. 27. H. Nourkhiz Mahjoub, A. Tahmasbi-Sarvestani, S. M. O. Gani, and Y. P. Fallah, “Composite α-μ Based DSRC Channel Model Using Large Data Set of RSSI Measurements,” in IEEE Transactions on Intelligent Transportation Systems. doi: 10.1109/TITS.2018.2803628 28. D. W. Matolak and J. Frolik, “Worse-than-Rayleigh fading: Experimental results and theoreti- cal models,” Commun. Mag. IEEE, vol. 49, no. 4, pp. 140–146, Apr. 2011. 29. M. Amro, A. Landolsi, S. Zummo, M. Grieger, M. Danneberg and G. Fettweis, “Ray-tracing wireless channel modeling and verification in Coordinated Multi-Point systems,” 2014 12th International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOpt), Hammamet, 2014, pp. 16–21. 30. CAMP VSC3 Consortium, “Interoperability Issues of Vehicle-to-Vehicle Based Safety Sys- tems Project (V2V-Interoperability) Phase 1 Final Report,” NHTSA Publication, 2014, (USDOT Docket ID: NHTSA-2014-0022-0029).
Applications of Connectivity in Automated Driving Ahmed Hamdi Sakr, Gaurav Bansal, Vladimeros Vladimerou, Kris Kusano, and Miles Johnson System Model In this chapter, we focus on solving two main problems in modern intelligent transportation systems; namely, localization and estimating the road geometry ahead of an ego vehicle, using on-board sensors (such as cameras, radars, etc.) and data received from neighboring vehicles via DSRC. In this context, road geometry is defined as the description of the center of the current lane of the ego vehicle. Sensor Setup The ego vehicle is equipped with a camera system that detects the left and right lane markings of the road and provides a description of the center of the lane by its relative offset ykoff and relative heading ϕk with respect to the ego vehicle, initial curvature c0,k, curvature change rate c1,k, and lane width wk at time k. In addition, a ranging sensor is installed on the ego vehicle to measure the relative position (xki , yki ) and velocity vki of different objects (including pedestrians, cyclists, vehicles, etc.) in its surroundings. The index i is a unique identification given by the ranging sensor to each detected object such that points belonging to the same object are clustered together, filtered, and classified. In practice, a ranging sensor can be a camera, a A. H. Sakr ( ) · G. Bansal 245 Toyota InfoTechnology Center, USA, Mountain View, CA, USA e-mail: [email protected]; [email protected] V. Vladimerou · K. Kusano · M. Johnson Toyota Motor North America R&D, Ann Arbor, MI, USA e-mail: [email protected]; [email protected]; [email protected] © Springer Nature Switzerland AG 2019 R. Miucic (ed.), Connected Vehicles, Wireless Networks, https://doi.org/10.1007/978-3-319-94785-3_10
246 A. H. Sakr et al. radar, and/or a lidar where the accuracy of the measurements and the complexity of the algorithms involved in the aforementioned process vary from one type of sensor to another. For example, a camera (such as Mobileye) can have an accuracy of less than 2.25 m at 45 m and 9 m at 90 m [1], a radar (such as Delphi ESR) can have an average accuracy of less than 0.5 m and 0.5◦, and a lidar (such as Velodyne HDL-32E and HDL-64E) can have a range accuracy of less than 2 cm. Throughout the work, the ego vehicle is equipped with a forward-looking radar that detects the relative position information of leading vehicles. Furthermore, the ego vehicle and few remote vehicles (RVs) are also equipped with a GPS receiver, a speed and yaw rate sensor, and a DSRC transceiver. The GPS receiver and speed and yaw rate sensor detect the vehicle’s current position (xk, yk), heading φk, speed vk, and yaw rate ψk in a fixed global coordinate system. The DSRC transceiver enables the vehicles to exchange their own state information (e.g., via SAE DSRC BSM [2–4]) with other equipped vehicles in their vicinity. Typically, each DSRC-equipped vehicle is required to broadcast its current position information (such as latitude, longitude, speed, heading, path history, etc.) every time slot Ts which is typically 100 ms. Note that the accuracy of the position information highly depends on the environment and the GPS receiver quality. That is, a GPS receiver in general gives more precise readings in an open-sky environment because of availability of more satellites as compared to urban environment that can suffer from severe signal propagation conditions [5]. Hereinafter, leading vehicles that are equipped with a DSRC transceiver are referred to as equipped remote vehicles (ERV) while vehicles with no on-board DSRC transceivers are referred to as non-equipped remote vehicles (NRV). Figure 1 shows a typical scenario with an ego vehicle and a number of remote vehicles. In this scenario, the ego vehicle receives BSM messages that are transmitted by nearby ERVs containing their position information over DSRC links (i.e., dashed red lines in Fig. 1). Upon receiving these messages, the ego vehicle extracts the position information and keeps track of each ERV’s position using a off 0φ Fig. 1 Basic scenario with a three-lane road. Dashed blue lines represent radar signals while dashed red lines represent DSRC links. Dashed black lines represent lane markings and solid black line represents the center of the lane of the ego vehicle
Applications of Connectivity in Automated Driving 247 unique identifier (e.g., a MAC address) given to each vehicle. Moreover, the ego vehicle uses the on-board radar to detect nearby RVs and measure their relative distances and angles (i.e., dashed blue lines in Fig. 1). The radar system also creates a unique identifier to keep track of range measurements belonging to the same RV using its own unique identifier. It is worth mentioning that the unique identifier given to each detected RV by the radar is not necessarily the same unique identifier used to keep track of the received BSM messages since both systems operate independently. Note also that an ego vehicle does not necessarily receive BSM messages from all vehicles detected by the radar which could happen due to either bad propagation conditions or that vehicle is an NRV. For example, in Fig. 1, although the vehicle in the middle lane ahead of the ego vehicle is an NRV and does not communicate with the ego vehicle, it is detected by the radar. One the other hand, the ERV in the right lane ahead of the ego vehicle is connected to the ego vehicle but not detected by the radar. Furthermore, the ERV in the left lane ahead of the ego vehicle is detected by the radar sensor and connected to the ego vehicle via DSRC at the same time. Hence, the number of RVs detected by the radar sensor is not necessarily the same number of RVs detected by the DSRC transceiver. Hereinafter, we use D = {ERV1, ERV2, . . . , ERVM } to denote the set of ERVs whose BSM messages are received by the DSRC transceiver at the ego vehicle where ERVi is a unique identifier given to the DSRC messages received from the same ERV. Also, we use R = {RV1, RV2, . . . , RVK } to denote the set of RVs detected by the radar sensor at the ego vehicle where RVj is a unique identifier given to each detected RV which could be either an ERV or NRV. Figure 1 also shows the lane markings measurements, i.e., the initial curvature c0 and curvature change rate c1 of the center of the ego vehicle’s lane, and relative offset yoff and relative heading ϕ of the ego vehicle with respect to the center of its lane. State Vector Representation First, we define the global state vector sk that contains the state of the ego vehicle, road geometry, and RVs. The vector sk has four main parts and is expressed as ⎡xke ⎤ (1) sk = ⎢⎣⎢xxkkdr ⎥⎥⎦ rk where xek is the state vector of the ego vehicle, xkr is the state vector of K RVs that are detected by the radar sensor, xdk denotes the state of M ERVs that are communicating with the ego vehicle via DSRC, and rk is the state vector of the road.
248 A. H. Sakr et al. Ego Vehicles The ego vehicle state vector is expressed as (2) ⎡⎤ xke = ⎢⎣⎢⎢⎢⎢φyxvkkkk ⎥⎥⎥⎥⎦⎥ ψk where xk and yk are the location of the ego vehicle in the global Cartesian coordinate systems in m, vk is the speed in m/s, φk is the heading in rad, and ψk is the yaw rate in rad/s. Remote Vehicles The vector xrk which contains the state vectors of RVs that are detected be the radar sensor is expressed as ⎡⎤ ⎢⎡⎢⎢⎣yxx˙kkkrrr,,,iii ⎤ xkr,1 ⎥⎥⎦⎥ xkr = ⎢⎢⎢⎢⎣ xkr,2 ⎦⎥⎥⎥⎥ , xkr,i = . (3) ... y˙kr,i xkr,K where K is the total number of detected RVs and xrk,i is the state vector of the i-th RV. Similarly, the vector xkd which contains the state vectors of ERVs that are detected be the DSRC transceiver is expressed as ⎡⎤ ⎡ xkd,i ⎤ xkd,1 ⎣⎢⎢⎢⎢⎢ ykd,i ⎥⎦⎥⎥⎥⎥ xdk = ⎢⎢⎣⎢⎢ xdk,2 ⎥⎦⎥⎥⎥ , xkd,i = vkd,i . (4) φkd,i ... xdk,M ψkd,i where M is the total number of detected ERVs and xkd,i is the state vector of the i-th ERV. Road Geometry As stated earlier, the road is defined as the center of the lane of the ego vehicle. We use a clothoid-based model to describe the road geometry ahead of the ego vehicle [6, 7]. The road is split into N segments connected end to end as shown in Fig. 2 where the length of the n-th segment is ln. The geometry of each segment is described by its curvature. We assume that the curvature cn(s) of a segment n changes linearly with the distance s along the road (i.e., the arc length). Hence, cn(s) = c0n + c1ns, s ∈ [0, ln] (5)
Applications of Connectivity in Automated Driving 249 where c0n and c1n are the initial curvature and curvature change rate of the n-th segment. By knowing the position of the ego vehicle relative to the first point of the first segment, the road geometry can be fully described at any time k by the following road state vector rk = ⎢⎢⎢⎢⎢⎡⎢⎢⎢⎣ccyϕ101ko...k,,fkkf ⎤ (6) ⎥⎥⎥⎥⎥⎦⎥⎥⎥ c1N,k where ykoff is the lateral offset between the ego vehicle and the center of its lane, and ϕk is the heading of the starting point of the road (i.e., segment 1) relative to the ego vehicle. Note that c0,k is the initial curvature of the first segment of the road which starts at the position of the ego vehicle at time k. This representation assumes that the road is continuous between each two consecutive segments (as shown in Fig. 2). As a result, each segment can be represented in the Cartesian coordinate system (xn(s), yn(s)) at distance s along the road as s (7) xn(s) = x0n + cos ϕn(t) dt 0 s yn(s) = y0n + sin ϕn(t) dt 0 where (x0n, y0n) is the starting point of the n-th segment and ϕn(s) is the heading of the segment which can be derived by integrating the curvature in Eq. (5) as follows ϕn(s) = ϕ0n + c0ns + 1 c1ns 2 (8) 2 where ϕ0n is the initial heading of the n-th segment. off 22 1 1φ 2 3 4 0 0 0 0 Fig. 2 Road model with N = 4 segments. The dots represent both ends of each segment and dashed lines represent the lane markings. The parameters c0n, c1n, and ln are the initial curvature, curvature change rate, and length of the n-th segment. The parameters yoff and ϕ are the lateral offset and relative heading of the ego vehicle
Search
Read the Text Version
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 20
- 21
- 22
- 23
- 24
- 25
- 26
- 27
- 28
- 29
- 30
- 31
- 32
- 33
- 34
- 35
- 36
- 37
- 38
- 39
- 40
- 41
- 42
- 43
- 44
- 45
- 46
- 47
- 48
- 49
- 50
- 51
- 52
- 53
- 54
- 55
- 56
- 57
- 58
- 59
- 60
- 61
- 62
- 63
- 64
- 65
- 66
- 67
- 68
- 69
- 70
- 71
- 72
- 73
- 74
- 75
- 76
- 77
- 78
- 79
- 80
- 81
- 82
- 83
- 84
- 85
- 86
- 87
- 88
- 89
- 90
- 91
- 92
- 93
- 94
- 95
- 96
- 97
- 98
- 99
- 100
- 101
- 102
- 103
- 104
- 105
- 106
- 107
- 108
- 109
- 110
- 111
- 112
- 113
- 114
- 115
- 116
- 117
- 118
- 119
- 120
- 121
- 122
- 123
- 124
- 125
- 126
- 127
- 128
- 129
- 130
- 131
- 132
- 133
- 134
- 135
- 136
- 137
- 138
- 139
- 140
- 141
- 142
- 143
- 144
- 145
- 146
- 147
- 148
- 149
- 150
- 151
- 152
- 153
- 154
- 155
- 156
- 157
- 158
- 159
- 160
- 161
- 162
- 163
- 164
- 165
- 166
- 167
- 168
- 169
- 170
- 171
- 172
- 173
- 174
- 175
- 176
- 177
- 178
- 179
- 180
- 181
- 182
- 183
- 184
- 185
- 186
- 187
- 188
- 189
- 190
- 191
- 192
- 193
- 194
- 195
- 196
- 197
- 198
- 199
- 200
- 201
- 202
- 203
- 204
- 205
- 206
- 207
- 208
- 209
- 210
- 211
- 212
- 213
- 214
- 215
- 216
- 217
- 218
- 219
- 220
- 221
- 222
- 223
- 224
- 225
- 226
- 227
- 228
- 229
- 230
- 231
- 232
- 233
- 234
- 235
- 236
- 237
- 238
- 239
- 240
- 241
- 242
- 243
- 244
- 245
- 246
- 247
- 248
- 249
- 250
- 251
- 252
- 253
- 254
- 255
- 256
- 257
- 258
- 259
- 260
- 261
- 262
- 263
- 264
- 265
- 266
- 267
- 268
- 269
- 270
- 271
- 272
