Development of potential methods for testing congestion control algorithm implemented in vehicle-to-vehicle communications

ABSTRACT

Objective: A channel congestion problem might occur when the traffic density increases because the number of basic safety messages carried on the communication channel also increases in vehicle-to-vehicle communications. A remedy algorithm proposed in SAE J2945/1 is designed to address the channel congestion issue by decreasing transmission frequency and radiated power. This study is to develop potential test procedures for evaluating or validating the congestion control algorithm.

Methods: Simulations of a reference unit transmitting at a higher frequency are implemented to emulate a number of onboard equipment (OBE) transmitting at the normal interval of 100 ms (10 Hz). When the transmitting interval is reduced to 1.25 ms (800 Hz), the reference unit emulates 80 vehicles transmitting at 10 Hz. By increasing the number of reference units transmitting at 800 Hz in the simulations, the corresponding channel busy percentages are obtained. An algorithm for Global Positioning System (GPS) data generation of virtual vehicles is developed for facilitating the validation of transmission intervals in the congestion control algorithm.

Results: Channel busy percentage is the channel busy time over a specified period of time. Three or 4 reference units are needed to generate channel busy percentages between 50 and 80%, and 5 reference units can generate channel busy percentages above 80%. The proposed test procedures can verify the operation of congestion control algorithm when channel busy percentages are between 50 and 80% and above 80%. By using a GPS data generation algorithm, the test procedures can also verify the transmission intervals when traffic densities are 80 and 200 vehicles in a radius of 100 m. A suite of test tools with functional requirements is also proposed for facilitating the implementation of test procedures.

Conclusions: The potential test procedures for a congestion control algorithm are developed based on the simulation results of channel busy percentage and the GPS data generation algorithm. The test tools can examine the requirement compliance automatically and objectively. The required number of reference units should be further validated using real OBEs before implementing these potential test procedures.

KEYWORDS:

• Vehicle-to-vehicle communication
• congestion control algorithm
• channel busy percentage
• test procedures

Introduction

Vehicle-to-vehicle (V2V) safety applications including emergency electronic brake lights, forward collision warning, blind spot/lane change warning, intersection movement assist, left turn assist, and control loss warning are potentially able to reduce traffic fatalities and injuries in the future. Safety applications of connected vehicles utilize V2V communications to exchange basic safety messages (BSM) for providing onboard warnings to drivers when impending crashes are detected. BSMs are transmitted 10 times per second to satisfy the short latency requirement of all safety applications. The BSM transmission latency should be short preferably; however, when traffic density increases, some BSMs may be lost for the sake of channel congestion. The excessive BSMs from several hundreds of vehicles cause a congestion problem on the communication channel. This congestion problem of V2V communications can be mitigated theoretically by reducing the transmission frequency and power. Reducing transmission frequency can decrease the number of transmitted BSMs in a period of time, and reducing

radiated power can decrease the transmission range such that fewer vehicles may sense transmitted BSM signals on the channel. The congestion control algorithm (CCA) proposed by SAE J2945/1 (SAE International 2015) specifies the details of how to reduce the transmission frequency and power based on the traffic density and channel busy percentage (CBP). The ultimate goal of this research is to develop the test procedures for evaluating the compliance of CCA performance on the vehicle under test. The research described in this article is conceptual and presents potential methods for evaluating or validating (spot-checking) the CCA. With some additional research and development, the authors believe that these concepts and methods have the potential to be developed into objective test procedures.

CBP is calculated by dividing the channel busy time by the specified total time of measurement (e.g., 100 ms). The time of either transmission or reception is deemed channel busy time. Reducing the radiated power from the normal level (20 dBm) to a lower level (10 dBm) will decrease the transmission range and result in fewer vehicles receiving BSM signals. For those vehicles out of the transmission range, their CBPs should be lower than those within the transmission range. When a CBP remains at a lower level (50% or less), the pretransmission waiting time (the transmitter needs to wait until the channel is clear to send BSMs) would be shorter. In addition, the probability of 2 packets arriving simultaneously (causing a collision) at the receiver end would be lower. When a packet collision occurs, the receiver is unable to receive the BSM correctly. Therefore, a higher CBP will cause a longer time to transmit and a higher packet error ratio (PER). PER is defined as the number of lost BSMs over the total number of BSMs that an on-board equipment (OBE) has expected from a remote OBE in the dedicated short range communications (DSRC) channel.

A simulation tool that is able to reproduce CCA operations with an adequate degree of fidelity can be used in the development of CCA test procedures. To ensure a minimum level of performance, the potential test procedures should be able to simulate a congestion environment in terms of CBP and traffic density using a minimum number of communication devices. These reference devices are capable of transmitting and receiving BSMs as OBEs installed in vehicles. In addition, they are able to transmit BSMs at a higher frequency (>10 Hz) to emulate a number of vehicles transmitting BSMs. These devices may also be configured to simulate Global Positioning System (GPS) operations of a number of vehicles whose GPS coordinates are transmitted in their BSMs. The GPS data can be generated from an algorithm that creates a number of traveling virtual vehicles in the proximity of the vehicle under test. The vehicle under test detects the channel busy status and traffic density generated from the reference test devices and starts to implement CCA that reduces transmitting power or frequency. The proposed test procedures can check the intervals between received BSMs and their corresponding radiated powers to determine the test result. Simulations of V2V communications can be used to decide the minimum number of reference test devices providing the required channel busy conditions. Based on the conditions of CBP and traffic density, potential test procedures are developed accordingly.

Related works

Channel congestion control of V2V safety applications can be achieved by adjusting the transmission frequency (or interval) and/or radiated power. Huang et al. (2010) proposed an algorithm that controls the transmission frequency and range based on estimated tracking errors and CBP. Bansal and Kenney (2013) defined an adaptive control algorithm, a weighted version of the linear message rate integrated control algorithm for the objective of maximizing channel throughput (messages/s). Linear message rate integrated control adjusts transmission frequencies proportionally as a function of CBP based on the difference between the goal and measured transmission frequencies. The European Telecommunications Standards Institute defines a reactive state-based approach of decentralized congestion control that also adjusts the transmission frequency as a function of CBP (Rostami et al. 2016). Fallah et al. (2011) defined theinformation dissemination rate (similar to channel throughput) by controlling transmission power based on CBP. Fallah et al. (2016) further improved their transmission power control algorithm to be stateful utilization-based power adaptation (SUPRA) that controls radiated power incrementally as a function of CBP. After reviewing related studies of CCA, it is found that CCA essentially adopts the linear memoryless range control algorithm (Huang et al. 2010) with the enhancement of considering traffic densities to control transmission frequencies. In addition, SUPRA is adopted to control radiated power. The goal of CCA is to control CBP and PER to an acceptable level for V2V safety applications. However, a maximum transmission interval of 600 ms defined in SAE J2945/1 may not be feasible for a forward collision warning application. A long information age (communication delay) can increase the risk of collision, especially when vehicle speeds are less than 40 mph (Hsu 2016). The main cause of this higher risk is due to the traveled distance during drivers' perception–reaction time. Automatic emergency braking may eliminate this human delay without compromising the safety requirement. Inman et al. (2016) also concluded that full cooperative adaptive cruise control (warning and automated braking) is needed to obtain a crash avoidance benefit. Overall, no test procedures of CCA have been discussed in the reviewed studies that focus on comparing the performance or channel stability of proposed algorithms by using simulations.

Methods

To develop the test procedures of CCA, the content of CCA is reviewed first. After reviewing CCA, it is found that V2V simulations are needed to figure out the needed number of reference OBEs for generating the required CBPs. In addition, for the purpose of generating vehicles automatically in the testing process, an innovative GPS data generation algorithm is proposed.

Congestion control algorithm

Figure 1 shows the flowchart of CCA operation as specified in SAE J2945/1. The CBP is defined as the percentage of time the channel is busy during a given interval (e.g., 100 ms). A channel is busy when the OBE radio is either transmitting or receiving. If a vehicle is moving from a location of low traffic density to a location of high traffic density and vice versa, then the vehicle's OBE may experience a sudden jump or drop of CBP. To avoid a fluctuation of CBP at every interval (100 ms), a moving average of raw CBPs is calculated to produce a smooth CBP. PERs are calculated between a pair of connected vehicles during a sliding window interval. BSM packets may be lost when an OBE is transmitting and a packet arrives or if 2 packets arrive at the same time. The channel quality indicator is calculated by averaging the PERs of all vehicles within a host vehicle's radius of 100 m. The host vehicle calculates the tracking error by comparing the difference of position estimation between the host and remote vehicles. The host vehicle estimates its current position based on the most recent GPS coordinate, speed, acceleration/deceleration, and heading. The remote vehicle's perception of the host vehicle position is

based on the most recent received BSM. If the remote vehicle receives BSMs from the host vehicle without any packet loss, then the tracking error should be small (the host vehicle's travel distance within 100 ms). Conversely, if BSM packets are lost several times, the tracking error can be large. Whether the remote vehicle can receive BSMs from the host vehicle correctly is determined by assessing the channel quality indicator. The channel quality indicator is the average of all remote vehicles' PERs in the radius of 100 m. The transmission probability is calculated based on the tracking error. When the tracking error exceeds the upper threshold (0.5 m; SAE International 2015), the transmission probability is 1. Otherwise, the transmission probability is between 0 and 1. If a drawn random number between 0 and 1 is less than the transmission probability for 3 consecutive times, the decision dynamics will trigger an immediate transmission.

Figure 1. CCA operation flowchart.

Maximum inter-transmit time (MaxITT) is calculated based on a smoothed traffic density at a radius of 100 m as denoted in Equation 1(1) $$N_s\left(k\right)=\lambda \times N\left(k\right)+(1-\lambda )\times N_s(k-1),$$(1) . The traffic density defined in SAE J2945/1 is the total number of vehicles in the radius of 100 m from the host vehicle. Equation 2(2) $$Max\_ITT\left(k\right)=\left\{\begin{array}{cc}100\hfill & N_s\left(k\right)\le B\\ 100\times \frac{N_s\left(k\right)}{B}\hfill & B<N_s\left(k\right)<\frac{vMax\_ITT}{100}\times B\\ vMax\_ITT\hfill & \frac{vMax\_ITT}{100}\times B\le N_s\left(k\right)\end{array}\right.$$(2) denotes the calculation of MaxITT from 100 to 600 ms. When the density is not greater than 25 vehicles, MaxITT is 100 ms, and when the density is not less than 150 vehicles, MaxITT is 600 ms. The parameters specified in Equation 2(2) $$Max\_ITT\left(k\right)=\left\{\begin{array}{cc}100\hfill & N_s\left(k\right)\le B\\ 100\times \frac{N_s\left(k\right)}{B}\hfill & B<N_s\left(k\right)<\frac{vMax\_ITT}{100}\times B\\ vMax\_ITT\hfill & \frac{vMax\_ITT}{100}\times B\le N_s\left(k\right)\end{array}\right.$$(2) are defined by SAE J2945/1. A host vehicle transmits BSMs at the interval of MaxITT, when no critical or decision dynamics event occurs. A critical event can be a sudden hard braking or any other safety related events. A decision dynamics event is triggered by tracking errors. If a critical or decision dynamics event occurs, a host vehicle will transmit a BSM immediately. (1) $$N_s\left(k\right)=\lambda \times N\left(k\right)+(1-\lambda )\times N_s(k-1),$$(1) where λ is the weight factor (0.05), and N_s(k) is the smoothed traffic density at time k. (2) $$Max\_ITT\left(k\right)=\left\{\begin{array}{cc}100\hfill & N_s\left(k\right)\le B\\ 100\times \frac{N_s\left(k\right)}{B}\hfill & B<N_s\left(k\right)<\frac{vMax\_ITT}{100}\times B\\ vMax\_ITT\hfill & \frac{vMax\_ITT}{100}\times B\le N_s\left(k\right)\end{array}\right.$$(2) where Max_ITT(k) is the message generation interval in milliseconds at time k, N_s(k) is traffic density at time k, B is the density coefficient (25), and vMax_ITT is the maximum threshold (600 ms).

The radiated power is determined by the CBP as shown in Equation 3(3) $$f\left(CBP\right)=\left\{\begin{array}{cc}vRPMax\hfill & CBP\le vMinChU\\ vRPMax-\left(\frac{vRPMax-vRPMin}{vMaxChU-vMinChU}\right)\times (CBP-vMinChU)\hfill & vMinChU<CBP<vMaxChU\\ vRPMin\hfill & vMaxChU\le CBP\end{array}\right.,$$(3) . The radiated power is between 10 and 20 dBm when a CBP is between 50 and 80%. Equation 4(4) $$\begin{array}{ccc}& & \mathrm{Base}\_\mathrm{RP}=\mathrm{Previous}\_\mathrm{RP}+\mathrm{vSUPRAGain}\times \left(\mathrm{f}\left(\mathrm{CBP}\right)\right.\hfill \\ & & \left.-\mathrm{Previous}\_\mathrm{RP}\right),\hfill \end{array}$$(4) denotes the current radiated power that is a smoothed result of the previous radiated power and a function of CBP. CCA calculates the current radiated power and determines MaxITT at each time interval, and when the scheduled BSM transmission interval expires (MaxITT) the OBE transmits a BSM with the current radiated power. (3) $$f\left(CBP\right)=\left\{\begin{array}{cc}vRPMax\hfill & CBP\le vMinChU\\ vRPMax-\left(\frac{vRPMax-vRPMin}{vMaxChU-vMinChU}\right)\times (CBP-vMinChU)\hfill & vMinChU<CBP<vMaxChU\\ vRPMin\hfill & vMaxChU\le CBP\end{array}\right.,$$(3) where vRPMax is the higher threshold for radiated power (20 dBm), vRPMin is the lower threshold for radiated power (10 dBm), vMinChU is the lower threshold for channel utilization (50%), and vMaxChU is the higher threshold for channel utilization (80%). (4) $$\begin{array}{ccc}& & \mathrm{Base}\_\mathrm{RP}=\mathrm{Previous}\_\mathrm{RP}+\mathrm{vSUPRAGain}\times \left(\mathrm{f}\left(\mathrm{CBP}\right)\right.\hfill \\ & & \left.-\mathrm{Previous}\_\mathrm{RP}\right),\hfill \end{array}$$(4) where RP is radiated power, and vSUPRAGain is the SUPRA gain (0.5).

Simulation of V2V communications

The purpose of utilizing the simulation software is to find the number of vehicles that may produce the DSRC channel busy environment needed for testing the CCA. V2V communications are emulated using VEhicles In Network Simulation (VEINS). VEINS is an open source vehicular network simulation framework for running vehicle network simulations. It is based on 2 well-established simulators: OMNeT++, a discrete event–based network simulator, and Simulation of Urban MObility, a road traffic simulator (Sommer et al. 2011). The Nakagami radio fading model is implemented to emulate the received signal power distribution. When the distance between 2 OBEs becomes longer, the received power weakens. DSRC power fading is modeled as a function of distance between 2 OBEs. The parameters of Nakagami distribution are calculated based on the distance, DSRC frequency, and antenna gains (Hafeez et al. 2013). The implementation of CCA is achieved by modifying the codes of physical layer (IEEE 802.11) and MAC layer (IEEE 1609.4) operations originally developed in VEINS. Simulation of Urban MObility provides coordinates of vehicle positions to the module in VEINS for calculating the distance between any 2 vehicles in the network.

Table 1^{Figure 2}lists all of the simulation parameters. The data rate is based on the standard of SAE J2945/1, and the BSM packet length is an average of packets with a full security certificate (longer) or certificate digest (shorter). The calculation of CBP is based on the definition of SAE J2945/1 that is a smooth CBP as shown in Equations 5(5) $$RawCBP\left(k\right)=\frac{(100\times \mathrm{Duration}\mathrm{Channel}\mathrm{Indicated}\mathrm{as}\mathrm{Busy})}{vCBPMeasInt}$$(5) and 6(6) $$CBP\left(k\right)=vCBPWeightFactor\times RawCBP\left(k\right)+(1-vCBPWeightFactor)\times CBP(k-1),$$(6) . The weight factor in Equation 6(6) $$CBP\left(k\right)=vCBPWeightFactor\times RawCBP\left(k\right)+(1-vCBPWeightFactor)\times CBP(k-1),$$(6) is used to adjust the weights of current and previous CBPs. A 0.5 weight factor is essentially treating the weights of current and previous CBPs equally. A channel is busy when an OBE is either transmitting or its clear channel assessment function indicates that the channel is busy, as defined in IEEE 802.11 (IEEE-SA Standards Board 2012). For the purpose of reducing the number of OBEs in testing CCA, an OBE is configured to transmit BSMs at the interval of 1.25 ms (800 Hz) in simulations. With the increase in transmission frequency from 10 to 800 Hz, each reference OBE can emulate 80 normal OBEs transmitting at 10 Hz. Four simulations are performed to emulate the channel congestion environment of 160, 240, 320, and 400 vehicles transmitting at 10 Hz. The distances between all OBEs are within 100 m and no packet loss is expected at this distance. (5) $$RawCBP\left(k\right)=\frac{(100\times \mathrm{Duration}\mathrm{Channel}\mathrm{Indicated}\mathrm{as}\mathrm{Busy})}{vCBPMeasInt}$$(5) (6) $$CBP\left(k\right)=vCBPWeightFactor\times RawCBP\left(k\right)+(1-vCBPWeightFactor)\times CBP(k-1),$$(6) where k is the time step, RawCBP is the percentage of time the channel was busy, vCBPMeasInt is the CBP measure instance (100 ms), and vCBPWeightFactor is 0.5.

Table 1. Simulation parameters.

Parameter	Value
Data channel frequency	5.86 GHz (channel 172)
Bandwidth	10 MHz
Data rate	6 Mbps
Radiated power	20 dBm
Received power sensitivity	−92 dBm
Transmission interval	1.25 ms
BSM packet length	250 bytes
Simulation period	30 s

A few simulation trials are performed to obtain the highest transmission frequency of a reference test device. It is found that 1.25 ms of transmission interval equivalent to 800 Hz is appropriate for emulating 80 vehicles. With this base transmission interval, simulations are performed to obtain CBPs of 2, 3, 4, and 5 reference test devices emulating 160, 240, 320, and 400 vehicles.

GPS data generation algorithm

The test of CCA's BSM generation intervals requires at least 25 vehicles, because the density coefficient is 25 vehicles as specified in Equation 2(2) $$Max\_ITT\left(k\right)=\left\{\begin{array}{cc}100\hfill & N_s\left(k\right)\le B\\ 100\times \frac{N_s\left(k\right)}{B}\hfill & B<N_s\left(k\right)<\frac{vMax\_ITT}{100}\times B\\ vMax\_ITT\hfill & \frac{vMax\_ITT}{100}\times B\le N_s\left(k\right)\end{array}\right.$$(2) . A maximum of 150 vehicles are needed for testing the maximum threshold (600 ms). Using 150 vehicles with OBEs to transmit BSMs may incur a cost problem. For the purpose of reducing testing costs, a reference test device needs to generate GPS data of a number of vehicles traveling around the vehicle under test. The trajectories of each virtual vehicle are converted into GPS data format in BSMs and transmitted to the vehicle under test. The vehicle under test would perceive a number of vehicles after receiving those BSMs, and the CCA should update the traffic density based on vehicle counts. The algorithm of GPS data generation is presented in the following.

Using the position of the vehicle under test as the base point, all of the relative positions of virtual vehicles can be generated based on the vehicle coordinated system. As illustrated in Figure 2, assuming that a reference vehicle is 100 m ahead of the vehicle under test and they are in the same lane, 10 virtual vehicles can be inserted at the interval of 10 m between vehicles. Likewise, 10 virtual vehicles can be generated behind the vehicle under test at the interval of 10 m as well. Ten vehicles (passenger cars) traveling in 100 m is possible with a small gap between any 2 vehicles. The virtual vehicle generation would like to emulate real traffic flows with more vehicles within 100 m. As specified in SAE J2945/1, the vehicle position reported in a BSM is a point projected onto the surface of the roadway plane with reference to the WGS-84 coordinate system. The 2-dimensional vehicle coordinate system (assuming that all vehicles are on the same elevation) as shown in Figure A1 (see online supplement) defines the heading direction as the X axis and the perpendicular right-hand direction as the Y axis. Therefore, the vehicle positions in the right-hand side lane can be obtained by setting the Y distance to +3 m and the left-hand side lane as −3 m in the vehicle coordinate system. The next step is to convert these positions into the GPS coordinate system and update them every 100 ms according to the speed of each vehicle.

Figure 2. Concept of GPS data generation.

The GPS data generation algorithm is listed as follows:

1.	Calculate the relative (reference) heading between the vehicle under test and reference vehicle using Equation 7.
2.	Add 1 vehicle every 10 m ahead of the vehicle under test (using Equations 8(8) $${\phi }^{\text{'}}=\frac{N}{f_2}+\phi \left[6pt\right]$$(8) and 9(9) $${\lambda }^{\text{'}}=\frac{E}{f_{3}cos\phi }+\lambda \left[6pt\right]$$(9) with X = 10n, Y = 0, n = 1–10).
3.	Add 1 vehicle every 10 m behind the vehicle under test (using Equations 7(7) $$\begin{array}{ccc}\hfill \theta & =& \mathrm{atan}2\left(sin\Delta \lambda cos\phi 1,cos\phi sin\phi 1\right.\hfill \\ & & \left.-sin\phi cos\phi 1cos\Delta \lambda \right),\hfill \end{array}$$(7) and 8(8) $${\phi }^{\text{'}}=\frac{N}{f_2}+\phi \left[6pt\right]$$(8) with X = −10n, Y = 0, n = 1–10).
4.	If a right-hand side lane is needed, repeat steps 2 and 3 with Y = 3 m.
5.	If a left-hand side lane is needed, repeat steps 2 and 3 with Y = −3 m.
6.	Update the vehicle positons using the vehicle speed and time interval (100 ms). (7) $$\begin{array}{ccc}\hfill \theta & =& \mathrm{atan}2\left(sin\Delta \lambda cos\phi 1,cos\phi sin\phi 1\right.\hfill \\ & & \left.-sin\phi cos\phi 1cos\Delta \lambda \right),\hfill \end{array}$$(7)

where atan2 is the arctangent of x and y coordinates, θ is the heading from the vehicle under test to the reference vehicle, ϕ is the latitude of the vehicle under test, λ is the longitude of the vehicle under test, ϕ1 is the latitude of the reference vehicle, and λ1 is the longitude of the reference vehicle, Δλ = λ − λ1. (8) $${\phi }^{\text{'}}=\frac{N}{f_2}+\phi \left[6pt\right]$$(8) (9) $${\lambda }^{\text{'}}=\frac{E}{f_{3}cos\phi }+\lambda \left[6pt\right]$$(9) (10) $$E=\mathrm{Y}\cos \theta +\mathrm{X}\sin \theta \left[6pt\right]$$(10) (11) $$N=\mathrm{X}\cos \theta -\mathrm{Y}\sin \theta \left[6pt\right]$$(11) (12) $$f_1=\sqrt{f\left(2-f\right)}\left[6pt\right]$$(12) (13) $$f_2=\frac{a\left(1-{f_1}^2\right)}{{\left(1-{f_1}^2\left({(sin\phi )}^2\right)\right)}^{1.5}}\left[6pt\right]$$(13) (14) $$f_3=\frac{a}{{\left(1-{f_1}^2\left({(sin\phi )}^2\right)\right)}^{0.5}},$$(14) where ϕ′ is New_latitude (rad), λ′ is New_longitude (rad), Y is Across_distance (m), X is Ahead_distance (m), a is 6,378,137 (semi-major axis of Earth), f is 0.003353 (flattening), f₁ is eccentricity, f₂ is the radius of Earth in the meridian, and f₃ is the radius of Earth in the prime vertical.

As illustrated in Figure A1 (see online supplement), the position update of a vehicle can be calculated as follows:

1.	Calculate ΔT = T′ − T (ms).
2.	Calculate the estimated distance traveled by the vehicle in ΔT, X = Speed (m/s) × ΔT/1,000.
3.	Y = 0 for the travel lane, Y = 3 for the right-hand side lane, Y = −3 for the left-hand side lane.
4.	Use Equations 8(8) $${\phi }^{\text{'}}=\frac{N}{f_2}+\phi \left[6pt\right]$$(8) and 9(9) $${\lambda }^{\text{'}}=\frac{E}{f_{3}cos\phi }+\lambda \left[6pt\right]$$(9) to find the vehicle's new position at time T′.

Simulation scenarios and results

Figure 3 shows the simulation results of minimum required reference test devices. Figure 3A shows that most CBPs are fewer than 50% when 2 reference test devices are transmitting at 800 Hz, equivalent to 160 vehicles. Figure 3B shows that most CBPs are between 55 and 65% when 3 reference test devices are transmitting at 800 Hz, equivalent to 240 vehicles. Figure 3C shows that most CBPs are between 65 and 75% when 4 reference test devices are transmitting at 800 Hz, equivalent to 320 vehicles. Figure 3D shows that most CBPs are above 80% when 5 reference test devices are transmitting at 800 Hz, equivalent to 400 vehicles. The CBP range of 400 vehicles is similar to the simulation results of CAMP's scalability testing (Bansal et al. 2015). Based on the simulation results, we may use 2 reference test devices to generate CBPs fewer than 50%, 3 or 4 reference test devices to generate CBPs between 50 and 80%, and 5 reference test devices to generate CBPs above 80%.

Figure 3. Comparisons of channel busy percentages for (A) 160, (B) 240, (C) 320, and (D) 400 vehicles.

Development of potential test procedures

The development of potential test procedures should be informed by the following factors:

• The test procedures must be objective, repeatable, and reproducible.

• The test procedures should be automated.

• The test procedures should be sufficiently flexible to allow testing of a variety of systems (e.g., integrated, aftermarket OBEs) installed in a light vehicle. Each system may be tested differently; however, the minimum performance levels remain unchanged from system to system.

• The test procedures should be able to be easily implemented in different locations as long as the required test tools and test environment are available.

Using a reference OBE to transmit at a higher frequency can render the needed CBPs for testing purposes, but the randomness of BSM arrivals at the receiver OBE may not be the same as many OBEs transmitting at 10 Hz. Uniform BSM arrivals cause less packet collisions as compared to random BSM arrivals. Less packet collisions result in lower packet error ratios. Nonetheless, PER is not used in the pass criteria of CCA, and radiated powers and intervals are required instead. Table 2 lists 4 test cases with pass criteria for developing their corresponding test procedures.

Table 2. Test cases of congestion control algorithm.

Case number	Test function	Test condition	Pass criteria
1	Radiated power	50% < CBP < 80%	10 < RP < 20 dBm
2	Radiated power	CBP ≥ 80%	RP = 10 dBm
3	Transmission interval	25 < vDensity < 150	100 < MaxITT < 600 ms
4	Transmission interval	vDensity ≥ 150	MaxITT = 600 ms

Test tools

The potential test procedures may utilize or require a suite of approved and validated test tools to support the data collection and analysis processes. The test tool suite as shown in Figure A2 (see online supplement) consists of the packet capture tool (PCT) on the BSM reception vehicle, a DSRC reference unit (DRU) for sending BSMs to the vehicle under test, and post processing software including packet analysis (PAS) and compliance table generation (CTG). A DRU is the reference test device capable of receiving and transmitting BSMs to support specific test scenarios (e.g., CCA). It can transmit BSMs at the 800 Hz frequency specified in the potential test procedures of CCA. The PCT captures DSRC packets on multiple channels simultaneously and time stamps each packet with the corresponding time it was received. It can also keep track of the radiated power in the header of each BSM packet. The PAS tool decodes captured BSMs and validates them against the SAE J2945/1constraints. This tool also has the capability to process sets of data (e.g., all packets captured during a test session) and provide an analysis report as well as listing any anomalies identified. It can also analyze the inter-packet (time) interval and radiated power of each received BSM. The CTG tool processes test results to produce the compliance tables as specified in the Appendix A of SAE J2945/1.

For the CCA testing, the PCT captures BSMs transmitted from the OBE of vehicle under test and the DRU. The PCT sends all captured BSMs to the PAS tool for calculating the percentages of BSMs with reduced radiated power and frequency. The CTG tool collects the analysis results of the PAS tool and generates the compliance report.

The PAS tool should

• calculate the number of BSMs with radiated power greater than 10 and less than 20 dBm and the total number of BSMs in the range 50% < CBP < 80%.

• calculate the number of BSMs with radiated power equal to 10 dBm and the total number of BSMs with CBP ≥ 80%.

• calculate the percentage of BSMs with reduced radiated power > 95%.

• calculate the number of BSMs with 600 > MaxITT > 100 and the total number of BSMs with vDensity = 80.

• calculate the number of BSMs with 600 = MaxITT and the total number of BSMs with vDensity = 200.

• calculate the percentage of BSMs with reduced transmission frequency > 95%.

Potential test procedures

The upper part of Table 3 lists the test procedures for 50% < CBP < 80% and with a traffic density greater than 25 and less than 150 vehicles in a 100-m radius. The lower part of Table 3 lists the test procedures for CBP > 80% and with a traffic density greater than 150 vehicles in a 100-m radius. Four DRUs transmitting at 800 Hz can provide CBPs around 70% and 80 vehicles per 100 m that should trigger the vehicle under test to transmit at 13.3 dBm and an interval of 320 ms. Five DRUs transmitting at 800 Hz can provide CBPs around 85% and 200 vehicles per 100 m that should trigger the vehicle under test to transmit at 10 dBm and an interval of 600 ms. The execution time for each run of the test procedure 1 or 2 should be less than 10 min.

Table 3. Test procedures 1 and 2.

Procedure CCA #1
Purpose	Test CCA operations (test cases #1 and #3)
Note	For creating the channel congestion condition, 4 DRUs (DSRC reference unit) are configured to transmit BSMs at 800 Hz to emulate CBP greater than 50% and less than 80%. Each DRU generates 20 different vehicles' GPS data to emulate traffic densities of 80 vehicles
Tools	PCT, 4 DRUs
Setup	Four DRUs are located at 0°, 90°, 180°, and 270° around the vehicle under test within 100-m radius. The PCT is located less than 1 m from the vehicle under test
Run	Each DRU transmits BSMs at 800 Hz and generate 20 different vehicles' GPS data within 100-m radius from the vehicle under test for 5 min, and the start and end times of this run are recorded (80 vehicles)
Procedure CCA # 2
Purpose	Test CCA operations (test cases #2 and #4)
Note	For creating the channel congestion condition, 5 DRUs (DSRC reference unit) are configured to transmit BSMs at 800 Hz to emulate CBP greater than 80%. Each DRU generates 40 different vehicles' GPS data to emulate traffic densities of 200 vehicles
Tools	PCT, 5 DRUs
Setup	Five DRUs are located at 0°, 45°, 90°, 180°, and 270° around the vehicle under test within 100 m radius. The PCT is located less than 1 m from the vehicle under test
Run	Each DRU transmits BSMs at 800 Hz and generates 40 different vehicles' GPS data within 100-m radius from the vehicle under test for 5 min, the start and end times of this run are recorded (200 vehicles)

Based on the pass criteria listed in Table 2, it is expected that all BSMs transmitted from the vehicle under test should meet the criteria. However, considering all of the unexpected factors that may result in an anomaly of BSM generation and transmission, a small error tolerance can be adopted in the pass criteria. Therefore, a 5% error tolerance is suggested for the pass criteria. A small error tolerance is also considered for the radiated power (0.5 dBm) and transmission interval (±5 ms). The pass criteria of test procedure 1 are more than 95% of BSMs with the radiated power between 10 to 13.8 dBm and transmission intervals between 315 and 325 ms. The pass criteria of test procedure 2 are more than 95% of BSMs with the radiated power no greater than 10.5 dBm and transmission intervals between 595 and 605 ms.

The potential test procedures of CCA are developed to verify a vehicle under test can reduce radiated powers and increase transmission intervals given appropriate CBP and traffic density conditions. The simulations of V2V communications are used to calculate the required number of DRUs transmitting at 800 Hz. At least 4 DRUs may be needed to test the CCA performance based on the simulation results. The GPS data generation algorithm can be used to generate the required number of vehicles transmitting BSMs for facilitating the test procedures without using real vehicles. The proposed test tools and their functional requirements can be further developed into a suite of test software for automating and standardizing the test procedures.

The required number of DRUs should be further validated using DSRC devices before implementing these potential test procedures. Several test runs of the proposed test procedures should be conducted to discover and solve all unexpected problems. The objective test procedures will be specific about the needed equipment, captured data, postprocessed data, and compliance requirements.