RFP Attachment 8.pdf

Draft Report Submitted Under Cooperative Agreement DTFH61-07-H-00038

Impacts of Cooperative Adaptive Cruise Control on Freeway Traffic Flow

Dongyan Su, Steven E. Shladover, Xiao-Yun Lu and Christopher Nowakowski

California PATH Program Institute of Transportation Studies University of California, Berkeley

June 2011

Impacts of Cooperative Adaptive Cruise Control on Freeway Traffic

Flow

Abstract

This report describes the use of microscopic simulation to estimate the effect of varying market penetrations of ACC and CACC on highway capacity. The distribution of time gap settings that drivers from the general public used in a real field experiment were used in the simulation, making this the first study of the effects of ACC and CACC on traffic to be based on real data. The results show that the use of ACC is unlikely to change lane capacity significantly. However, CACC is able to greatly increase capacity after its market penetration reaches moderate to high percentages. The capacity increase can be accelerated by equipping non-ACC vehicles with “Here I Am” Onboard Units so that they can serve as the lead vehicles for CACC vehicles.

1. Introduction

An earlier phase of this project included a field test of Adaptive Cruise Control (ACC) and Cooperative Adaptive Cruise Control (CACC), driven by 16 drivers from the general public. Those drivers were encouraged to select the time gap settings that they preferred for each system, and their selections of time gap were recorded, along with many other parameters, for subsequent analysis. They were also surveyed to determine their subjective opinions about the ACC and CACC systems. The results of this experiment were very encouraging about the potential market acceptance of ACC and CACC when they are made available to the general public. These results were reported in two prior project reports (1, 2) and two technical papers (3, 4).

The C/ACC field test produced quantitative results indicating the relative preferences of the driving population for driving at the different available time gap settings. These time gap preferences can have a significant influence on traffic flow and highway lane capacity. The current work reported here uses a traffic microsimulation, combined with the C/ACC field test results, to produce the first authoritative quantitative estimates of the impacts that these systems could have on highway capacity.

The maximum traffic flow is determined by admissible time gaps between vehicles. The admissible time gap is determined by the means of controlling the vehicle’s car following: manual driving, Adaptive Cruise Control, or Cooperative Adaptive Cruise Control. In manual driving, the acceptable time gap is determined based on the driver’s perception of what is safe, including his or her perception and reaction time, and is influenced by the driver’s experiences, including expectations about the behaviors of other drivers, especially the driver of the leading vehicle. Vehicles with ACC or CACC have discrete time gap settings, which the driver can select based on his or her perceptions of the capabilities of the system. The net effect on traffic depends on whether drivers have sufficient confidence in the C/ACC systems to select time gaps that differ significantly from the gaps they use in manual driving.

This research evaluates the effects of the use of ACC and CACC on freeway capacity by microscopic simulation, based on the actual gaps that the drivers selected in our field testing of these systems. The simulation platform for this study is the commercially available traffic microsimulation program Aimsun, which was selected because it was the only simulation platform in which we could implement the NGSIM over-saturated freeway flow model, to provide the most realistic representation of normal drivers’ car following and lane changing behavior in dense traffic.

2. Vehicle Types to be Simulated

There are four vehicle types represented in the simulation, to accommodate all possible combinations of vehicles that could be interacting with each other in ways that would influence freeway traffic flow and capacity:

(a) Manual vehicle - driven manually by a driver, with car following behavior represented by the NGSIM oversaturated flow model

(b) Adaptive cruise control (ACC) - car following is determined based on a simple first-order control law representing the behavior of a typical ACC system, with relatively slow, gentle responses to changes by the car ahead.

(c) Here I am! (HIA) - driven manually just like the Manual vehicle, but it is equipped with a DSRC radio that frequently broadcasts a “here I am” message giving its location and speed. If it is being followed by a CACC vehicle, that following vehicle can use

CACC.

(d) Cooperative adaptive cruise control (CACC) - if it is following an HIA vehicle or another CACC vehicle, it can use its CACC car-following capability. If it is following a manual vehicle or an ACC vehicle, it acts like another ACC vehicle. The CACC car-following capability includes a faster response to changes by the car ahead and permits following at significantly shorter time gaps.

3. The Microsimulation Platform

The simulation is done on the platform of Aimsun, which is a transportation simulation environment developed by TSS (5). It contains microscopic and mesoscopic simulators, dynamic traffic simulator, macroscopic and static assignment models. It also offers extended tools for advanced investigation, like the Aimsun SDK (Software Development Kit), Aimsun API, and Aimsun MicroSDK. We used the Aimsun Microscopic Simulator, API, and microSDK in this project.

The Aimsun Microscopic Simulator (6) is the tool to construct traffic networks, define vehicle types and their basic properties, specify traffic demand and traffic control, and run the simulations. In this project, the geometry of the freeway and its speed limit, the four vehicle types and their properties, such as length and width, are defined in the Aimsun Microscopic Simulator.

The Aimsun API module (7) is an interface that allows external applications to access the internal data of Aimsun during simulation. The user can obtain all of the information during simulation, like the measurements of a detector or the state of a particular vehicle.

The user can also control the traffic, like determining when and how a new vehicle enters the network. Thus Aimsun API is used to record the measurements of traffic flow and to control the entering of new vehicles to the network during simulation.

The Microscopic Model SDK (8) is the tool to implement new behavior models in the Aimsun simulation, replacing the default driver model in Aimsun. The new driver behavior is implemented by the plug-in, which is a DLL file generated after building some C++ files. During each simulation step, Aimsun calls the functions in the plug-in and updates the driver behavior based on the user-defined model. So the new detailed driver behavior models that we had to develop to represent manual driving, ACC and CACC vehicles were programmed in MicroSDK.

4. Control Algorithm of ACC/CACC Vehicles

The following variables are used to define the vehicle-following control algorithms for ACC or CACC vehicles in this section. It should be noted that these are simplified representations of the ACC and CACC car-following rules that were actually implemented on the test vehicles for the field test. The ACC car-following rules are proprietary to Nissan, while the CACC car following behavior has been described in the technical paper by Bu, Tan and Huang (9). Simpler representations were needed here for computational efficiency, because they need to be executed many times in each simulation, and also because the finer details of the actual car following dynamics of these systems were implemented for driver comfort but probably have little influence on traffic flow dynamics.

v : the speed of the controlled ACC/CACC vehicle (m/s).

dv : the desired speed set by the driver, or the speed limit of the road (m/s).

ev : the speed error (m/s).

sca : the acceleration by speed control (m2/s).

s: the spacing between the controlled vehicle and its leading vehicle (m).

ds : the desired spacing (m).

es : the spacing error (m).

dh : the desired time gap (s).

Basically ACC and CACC vehicles have the same control algorithm. The only difference is they have different desired time gaps. There are two modes, speed control and gap control, in the ACC/CACC control algorithm. The goal of speed control is to keep the vehicle speed close to the speed limit, and that of gap control is to maintain the gap between the controlled vehicle and its leading vehicle to be the desired gap. Speed control is activated when the spacing to the preceding vehicle in the same lane is larger than 120 meters, and gap control is activated when the spacing is smaller than 100 meters.

If the spacing is between 100 meters and 120 meters, the controlled vehicle retains the previous control strategy to provide hysteresis to avoid dithering between the two strategies.

In speed control, the control law is sc esc de aa vbounda vvv

)2,2,4.0(

Where the function ()bound is defined as )),,max(min(:),,( lbublbub xxxxxxbound = . The values +2 and -2 are the maximum acceleration and deceleration of the vehicle under C/ACC control. This control law tries to eliminate the error between the vehicle speed and the speed limit of the road if the vehicle is in the speed control mode.

In gap control, the control law is

)2,,25.0(

)2,2,4.0( sce de dd esc de assbounda sss vhs vbounda vvv

The +2 and -2 here have the same meaning as in speed control. This control law forces the vehicle to approach its desired time gap set point in gap control. But the vehicle will still obey the speed limit in gap control, because if the commanded vehicle speed is larger than the speed limit, this law asks the vehicle to slow down, even if the current gap is larger than its desired gap.

5. Manual Driving Model

The following variables are used to explain the manual driving model in this section.

U nx and L nx : the upper bound and lower bounds for the driving distance (m).

τ : the wave travel time (s).

jam ng : the jam gap (m).

l : the length of the vehicle (m).

a : the acceleration of the vehicle (m2/s).

v : the speed of the vehicle (m/s).

x : the position of the vehicle (m).

fv : the free flow speed (m/s).

The manual driver behavior model is the NGSIM oversaturated freeway flow model developed by Yeo (10,11). The NGSIM oversaturated freeway flow model contains the car-following model and lane-changing model. But we only use the car-following model, the CF mode in (10,11), because there is no lane changing in the simulation. The basic car-following model of this model comes from Newell's linear model. It can be described as follows:

L n n n n jam nnnn

L nn

L nn

L n s n n L nnn

L n s nn f nn

U nnn jam nnnn

U n

L n

U nn a tv td tdgltxtxaaatttx txtattvtxttx ttxtxtvtxtattvtxglttxttx ttxttxttx

)()()()((2)()(

)}(,)()(max{)(

)}()(,)(,)()(,)(min{)(

)}(),(max{)( ττ τ

This NGSIM oversaturated freeway flow model was calibrated by using the NGSIM data (12).

6. A Freeway Section Model with Simplified Road Geometry

The tested road is a one-lane straight freeway with speed limit of 120 km/h (73 mph).

The freeway is 6.5 km long, and there is a detector located 6 km from the entrance. This location is selected to make sure all the flow measurements are in steady state. The freeway is empty before the simulation. During the simulation, the entering of new vehicles is controlled by the algorithm written in the API file, which will be described in the next paragraph. The total simulation length is 1 hour, and the simulation step is 0.1 second. The flow is recorded at intervals of 5 minutes, but the first measurement is discarded because the first 5 minutes are viewed as the warm-up time. The capacity is the average flow over the remaining 55 minutes.

The four types of vehicles have the same physical characteristics. The length is 4.7 meters and width is 1.9 meters. The maximum acceleration is 2 m/s2 and maximum deceleration is -2 m/s2.

In the simulation, the type of the next entering vehicle is randomly chosen, but follows the percentages defined in the simulation cases we want to test. The desired time gap of the entering vehicle is also random. For manual driving, the randomness is introduced by the randomness of jam ng .

We know that the maximum flow for manually driven vehicles on this type of simple freeway link should be about 2200 veh/h, so we assume the desired headway for manual driving is 1.64 sec (≈3600/2200). The desired time gaps of the ACC or CACC vehicles were selected based on the results of the field test (1-4), as shown in Figure 1:

ACC: 31.1% at 2.2 s time gap, 18.5% at 1.6 s time gap, 50.4% at 1.1 s time gap

CACC: 12% at 1.1 s time gap, 7% at 0.9 s time gap, 24% at 0.7 s time gap, 57% at 0.6 s time gap.

Note that the difference between headway and time gap needs to be accounted for by incorporating the incremental time needed to travel the vehicle length at the defined operating speed.

Figure 1 Distribution of time gap settings for CACC and ACC chosen by drivers while car following in field test

The desired entering headways for the ACC and CACC vehicles are chosen based on these time gaps, with the addition of the time increment to account for vehicle length.

The gap for manual driving is selected randomly during the simulation, within a +/- 10% error range of 1.64 s, that is, from 1.48 to 1.8 sec. At each simulation step, we check the travel time from the entrance to the location of the last entering vehicle, based on the speed of that vehicle at that step. If this travel time is larger than the desired entering time gap, we let a new vehicle enter the freeway at the same speed as its leading vehicle at that step. By this algorithm, the vehicles enter the freeway at an interval and speed that will not generate a measured maximum flow lower than the real capacity due to insufficient demand, while preventing vehicle collision because of entering at too high a speed or too small a time gap.

Because the entering time gap for manually driven vehicles usually does not match its desired time gap, the manually driven vehicles need to adjust their speeds after they enter the freeway. This causes the vehicles following them, whether they are manual, ACC or CACC, to also need to adjust speeds. By this, we introduce small disturbances into the simulation. This means that the measured maximum flow should be achievable and stable in traffic with small disturbances.

7. Simulation Scenarios and Results

Simulation scenarios have been defined to represent diverse combinations of manually driven, ACC, CACC and HIA vehicles so that the effects of changes in market penetration of each kind of vehicle can be determined. For each scenario, three simulations were run with different random number seeds and the results of those simulations were averaged to produce the estimates of achievable traffic flow.

The all-manual case was already referenced as a base case with a nominal capacity of 2200 veh/hr per lane. When basic ACC vehicles are incorporated into the traffic stream, the achievable traffic flow appears to be remarkably insensitive to the market penetration of ACC vehicles, as shown in Figure 2. Note that the flow remains within the narrow range from 2025 to 2080 vehicles per hour regardless of the market penetration. This is a consequence of the driver preferences for ACC time gap settings being very similar to the time gaps that they adopt when they drive manually. It is important to note that this refutes the assumptions in some published papers (13) that contend that ACC could substantially increase highway capacity.

Figure 2. Highway Lane Capacity (Vehicles/Hour) as a Function of Changes in ACC Market Penetration

If we consider only the combinations of manually driven and CACC vehicles, the trend in highway lane capacity with respect to CACC market penetration is as shown in Figure 3.

This has an obvious quadratic shape, based on the fact that the CACC vehicle can only use its CACC capability when it is following another CACC vehicle (or an HIA vehicle), but when it is following a manual vehicle it must revert to conventional ACC control. As a result of this, the capacity grows very slowly until the CACC market penetration becomes substantial, and then it grows much more rapidly. If all vehicles in a lane were equipped with CACC capability and the drivers chose the same distribution of CACC time gaps as they chose in our field test, the lane capacity would increase to 3970 vehicles per hour, a very dramatic improvement.

Figure 3. Highway Lane Capacity (Vehicles/Hour) as a Function of Changes in CACC Market Penetration Relative to Manually Driven Vehicles

One of the strategies being proposed in the Connected Vehicles initiative to improve performance of cooperative systems at low market penetrations is to equip as many existing vehicles as possible with a simple and inexpensive aftermarket positioning and communication Onboard Unit (OBU) that can broadcast a “Here I Am” (HIA) message.

This message provides the basic GPS coordinates and vehicle speed and heading information so that the OBUs on other vehicles can detect the trajectory of the vehicle.

This information, if it is sufficiently accurate, would enable and HIA equipped vehicle to be the leader for a CACC vehicle to follow at a short time gap. The effects of replacing the manually driven vehicles in the simulation cases of Figure 3 with HIA vehicles are shown in Figure 4. In this case, all the vehicles that do not have CACC are equipped with the HIA devices and can therefore serve as leaders for the CACC vehicles. With this change, the quadratic growth in Figure 3 becomes more nearly linear, and the capacity of the highway lane can be increased more significantly even at modest CACC market penetrations. At a 20% market penetration, the HIA addition increases capacity by 7%, at 30% market penetration it increases by more than 10% and in the 50% to 60% market penetration range the increase is in the range of 15% compared to the cases without HIA devices.

Figure 4. Highway Lane Capacity (Vehicles/Hour) as a Function of Changes in CACC Market Penetration Relative to “Here I Am” Vehicles

In our earlier studies of CACC, prior to the current project, we simulated the effects of the different combinations of ACC and CACC market penetrations, based on the assumption that the CACC vehicles would be driven at 0.5 s time gaps (14, 15). This produced a 3-D plot of achievable highway lane capacity that has been widely cited, and is reproduced here as Figure 5. The new simulation results, based on the time gaps that drivers actually chose in our field test, are shown in Figure 6 and Table 1. These capacity estimates are somewhat lower, with the 80% CACC/20% ACC result now in the range of 3000 rather than 3500 vehicles per hour, for example.

Figure 5 Original Prediction of Lane Capacity Effects of ACC and CACC Driven at 0.5 s Time Gap from 2001

Figure 6 Updated Prediction of Lane Capacity Effects of ACC and CACC Driven at Time Gaps Chosen by Drivers in Field Test (With the remaining vehicles manually driven)

Table 1 Updated Prediction of Lane Capacity Effects of ACC and CACC Driven at Time Gaps Chosen by Drivers in Field Test (With the remaining vehicles manually driven)

Percentage of CACC Vehicles 10% 20% 30% 40% 50% 60% 70% 80% 90%

10% 2054 2076 2149 2240 2363 2503 2703 2962 3382

20% 2050 2091 2152 2237 2349 2505 2671 2979 0

30% 2060 2102 2153 2243 2363 2486 2711 0 0

40% 2066 2108 2169 2249 2345 2525 0 0 0

50% 2073 2111 2164 2252 2362 0 0 0 0

60% 2080 2112 2157 2231 0 0 0 0 0

70% 2086 2109 2149 0 0 0 0 0 0

80% 2068 2112 0 0 0 0 0 0 0

P er ce nt ag e of A

C C

V eh ic le s

90% 2077 0 0 0 0 0 0 0 0

The capacity effects of different combination of CACC vehicles and HIA vehicles (with the rest being manually driven) are shown in Figure 7 and Table 2. As the market penetration of CACC increases, the increasing capacity attributable to the additional HIA vehicles can be seen, but it is a relatively subtle effect. For completeness, the analogous results for different combinations of CACC vehicles and HIA vehicles (with the rest being conventional ACC vehicles) are shown in Figure 8 and Table 3. Since the effects on capacity of ACC and manually driven vehicles are very similar, these results do not differ much from the previous results.

Figure 7 Prediction of Lane Capacity Effects of HIA and CACC Driven at Time Gaps Chosen by Drivers in Field Test (With the remaining vehicles manually driven)

Table 2 Prediction of Lane Capacity Effects of HIA and CACC Driven at Time Gaps Chosen by Drivers in Field Test (With the remaining vehicles manually driven)

10% 2105 2148 2187 2297 2459 2595 2847 3111 3624

20% 2109 2144 2175 2340 2469 2669 2956 3304 0

30% 2117 2163 2258 2384 2545 2789 3052 0 0

40% 2103 2175 2266 2400 2604 2904 0 0 0

50% 2115 2192 2292 2458 2699 0 0 0 0

60% 2114 2212 2338 2505 0 0 0 0 0

70% 2120 2222 2362 0 0 0 0 0 0

80% 2115 2231 0 0 0 0 0 0 0 P er ce nt ag e of H

IA

V eh ic le

90% 2114 0 0 0 0 0 0 0 0

Figure 8 Prediction of Lane Capacity Effects of HIA and CACC Driven at Time Gaps Chosen by Drivers in Field Test (With the remaining vehicles being ACC)

Table 3 Prediction of Lane Capacity Effects of HIA and CACC Driven at Time Gaps Chosen by Drivers in Field Test (With the remaining vehicles being ACC)

10% 2039 2102 2161 2264 2405 2600 2797 3152 3624

20% 2048 2111 2194 2296 2475 2672 2924 3304 0

30% 2059 2133 2222 2349 2546 2805 3052 0 0

40% 2066 2150 2257 2400 2628 2904 0 0 0

50% 2076 2171 2293 2463 2699 0 0 0 0

60% 2089 2188 2332 2505 0 0 0 0 0

70% 2095 2207 2362 0 0 0 0 0 0

80% 2105 2231 0 0 0 0 0 0 0 P er ce nt ag e of H

IA

V eh ic le

90% 2114 0 0 0 0 0 0 0 0

8. Concluding Remarks

The results reported here represent the first predictions of the effects of ACC and CACC on highway lane capacity that are founded on real experimental data, from drivers who have driven the suitably equipped vehicles and selected the time gap settings with which they were comfortable. These results show that conventional ACC is unlikely to produce any significant change in the capacity of highways, but CACC has the potential to substantially increase highway capacity when it reaches a moderate to high market penetration.

These results showed a maximum lane capacity of about 4000 vehicles per hour if all vehicles were equipped with CACC. If the vehicle population consists of CACC and HIA vehicles, meaning that all vehicles have been equipped with DSRC radios, the lane capacity increases approximately linearly from 2000 to 4000 as the percentage of CACC vehicles increases from zero to one hundred. On the other hand, if the vehicle population consists of manual and CACC vehicles, without any mandate for non-CACC vehicles to be equipped with DSRC, the increase in lane capacity follows a quadratic profile, lagging significantly behind at the intermediate market penetration values. Therefore, the capacity benefits of CACC can be accelerated, or obtained at somewhat lower market penetrations, if the rest of the vehicle population is equipped with HIA devices so that they can serve as the lead vehicles for the CACC vehicles.

References

1. S.E. Shladover, X.-Y. Lu, D. Cody, C. Nowakowski, Z. Qiu, A. Chow, J. O’Connell, J.

Nienhuis and D. Su, “Development and Evaluations of Selected Mobility Applications for VII”, PATH Research Report UCB-ITS-PRR-2010-25. Available online at:

http://database.path.berkeley.edu/reports/index.cgi?reqtype=displayrecord&record=1011

2. C. Nowakowski, S. E. Shladover, D. Cody, F. Bu, J. O’Connell, J. Spring, S. Dickey, and D. Nelson, “Cooperative Adaptive Cruise Control: Testing Drivers’ Choices of Following Distances”, PATH Research Report UCB-ITS-PRR-2011-01. Available online at:

http://database.path.berkeley.edu/reports/index.cgi?reqtype=displayrecord&record=1032

3. Steven E. Shladover, Christopher Nowakowski, Jessica O’Connell and Delphine Cody, “Cooperative Adaptive Cruise Control: Driver Selection of Car-Following Gaps”, ITS World Congress, Busan, Korea, November, 2010.

4. C. Nowakowski, D. Cody, J. O’Connell and S. E. Shladover, “Cooperative Adaptive Cruise Control: Driver Acceptance of Following Gap Settings Less Than One Second”, Human Factors and Ergonomics Society 54th Annual Meeting, San Francisco, September 2010.

5. “Aimsun Users Manual”, Transport Simulation Systems, July 2009.

6. “Microsimulator and Mesosimulator in Aimsun 6 User's Manual”, Transport Simulation Systems, April 2009.

7. “Aimsun Microsimulator API Manual”, Transport Simulation Systems, July 2009.

8. “Aimsun Microscopic Simulator Behavioural Models SDK, Transport Simulation

Systems”, May 2008.

9. F. Bu, H.-S. Tan and J, Huang, “Design and Field Testing of a Cooperative Adaptive Cruise Control System”, 2010 American Control Conference, Baltimore, MD.

10. Yeo, H. (2008). Asymmetric Microscopic Driving Behavior Theory, PhD thesis, University of California, Berkeley.

11. Yeo, H. et al. (2008). “Oversaturated freeway flow algorithm for use in next generation simulation”, Transportation Research Record, No. 2088, Transportation Research Board, Washington, D.C., pp. 68–79.

12. Yeo, H. and Skabardonis, A. (2008). “Parameter estimation for NGSIM freeway flow algorithm”, ASCE AATT 2008, Athens, Greece.

13. A. Kesting, M. Treiber, M, Schönhof and D. Helbing, “Adaptive Cruise Control Design for Active Congestion Avoidance”, Transportation Research Part C, Volume 16C, No. 6, Dec. 2008, pp. 668-683.

14. J. VanderWerf, S.E. Shladover, M.A. Miller and N. Kourjanskaia, “Effects of Adaptive Cruise Control Systems on Highway Traffic Flow Capacity”, Transportation Research Record No. 1800, Transportation Research Board, Washington DC, 2002, pp.

78-84.

15. Shladover, S; VanderWerf, J.; Miller, MA; Kourjanskaia, N; Krishnan, H, Development and Performance Evaluation of AVCSS Deployment Sequences to Advance from Today'sDriving Environment to Full Automation, California PATH Research Report UCB-ITS-PRR-2001-18. Available online at:

http://database.path.berkeley.edu/reports/index.cgi?reqtype=displayrecord&record=619