A cooperative driving system with automated vehicles and inter-vehicle by sparkunder20


									2001 IEEE Intelligent Transportation Systems Conference Proceedings - Oakland (CA) U S A = August 25-29, 2001

A Cooperative Driving System with Automated Vehicles and Inter-Vehicle Communications in Demo 2000
Sadayuki Tsugawa, Member, IEEE, Shin Kato, Kiyohito Tokuda, Takeshi Matsui, and Haruki Fujii
Abstract--This paper describes the technologies of the cooperative driving with automated vehicles and inter-vehicle communications in the Demo 2000 Cooperative driving. The cooperative driving, aiming at the compatibility of safety and efficiency of road traffic, here means that automated vehicles drive by forming a flexible platoon over a couple of lanes with a short inter-vehicle distance while performing lane changing, merging, and leaving the platoon. The vehicles for the demonstration are equipped with automated lateral and longitudinal control functions with localization data by the DGPS and the inter-vehicle communication function with 5.8 GHz DSRC designed for the dedicated use in the demonstration. In order to show the feasibility and potential of the technologies, the demonstration was held in November, 2000 on a test track with 5 automated vehicles. The scenario included stop & go, platooning, merging and obstacle avoidance.
Index Ternis-Advanced Vehicle Control and Safety Systems, Automated Highway Systems, Cooperative Driving, Inter-Vehicle Communications.

inter-vehicle communications linking the automated vehicles. which enables each vehicle to perform safe and efficient lane changing, merging and passing. The formation of the vehicles in the cooperative driving follows the wild geese and the dolphins. The formation of wild geese under the migration i:j aerodynamically efficient, and dolphins swim without collision while communicating with each other. The background of the cooperative driving presented here is the SSVS (Super Smart Vehicle System) studies from 1990 to 1993 [7] and the sbdies on the inter-vehicle communications with infrared from 1993 to 1997 [8]. In order to show the feasibility and the potential of the technologies for the cooperative driving, the demonstration “Demo 2000 cooperative driving” was conducted in November, 2000 on a test track. This paper describes the techno!ogies of the vehicle control [9] and the inter-vehicle communications with 5.8 GHz DSRC [IO] as well as the scenario of the demonstration [ 1 I].


provide essential solutions to the issues of the accidents, the congestion, the pollution and the energy consumption. Automated highway systems by not only a free agent but also a platoon have been demonstrated in the past several years [I]-[6]. The effect of the introduction of the automation on the safety can be investigated with an automated free agent, but that on the efficiency or the increase in the throughput of roadways must be investigated with an automated platoon. This paper describes the technologies for a flexible automated platooning or the cooperative driving. The cooperative driving, which is an advanced form of the automated highway systems, is defined here as flexible platooning of automated vehicles with a short inter-vehicle distance over a couple of lanes. It is featured by the use of the


HE introduction of automation into road traffic would

Under the steady state of the cooperative driving, the vehicles drive as shown in Fig. 1. On a single lane roadway the vehicles are arranged to make a string with a short inter-vehicle distance, and on a two lane or three latie roadways the vehicles are placed alternately and tightly over the lanes to make a flexible platoon, which enables smooth lane changing and merging. The inter-vehicle distance is d on a single lane roadway, and 1.5d on a multiple lane roadway. ‘I‘he formation of the flexible platoon will require inter-vehicle communications, because the onboard measurement of the locations, speeds, and acceleration of

Sadayuki Tsugawa is with the National Institute of Advanced Industrial Science and Technology, Namiki 1-2-1, fsukuba-shi, Ibaraki-ken, 305-8564 Japan (e-mail: tsugawa.s@aist.go.jp). Shin Kat0 is with the National Institute of Advanced Industrial Science and Technology, Namiki 1-2- 1, Tsukuba-shi, Ibaraki-ken, 305-8564 Japan (e-mail: shin.kato@aist.go.jp). Kiyohito Tokuda is with Oki Electric Industry Co., Ltd., Hikari-no-oka 3-4, Yokosuka-shi, Kanagawa-ken, 236-0847 Japan (e-mail: tokuda3 I5@oki.co.jp). Takeshi Matsui is with Denso Corporation, Showa-machi 1-1, Kariya-shi, Aichi-ken, 448-8661 Japan (e-mail: matsui@tao.denso.co.jp). Haruki Fujii is with Association of Electronic Technology for Automotive Traffic and Driving, Toranomon 34 Mori Bldg., Toranomon 1-25-5, Minato-ku, Tokyo, 105-0001 Japan (e-mail: fujii@jsk.or.jp).


Fig. 1. Cooperative driving formations under steady state.

0-7803-7194-1/01/$10.0002001 IEEE


2) Formation on two lane road
€ 7

An example of the process of forming the cooperative driving formation from an initial state to a platoon on a two lane roadway and a one lane roadway is shown in Fig. 2. At the initial state, 5 vehicles are parked side by side. After a while, a platoon under the steady state on a two lane roadway is formed. When one lane merges into the other, the vehicles on the bottom lane merge to the platoon on the top lane altemately. After driving on a single lane roadway, when a new lane appears, the platoon forms a new platoon over two lanes, and forms a new formation. A simulation study shows the feasibility of the process of the formation [9].

3 ) Merging

4) Formation on one lane road

e e

5 ) Separation onto two lane road

6 ) Forination on two lane road Fig. 2. Illustrative process of formation update.

The cooperative driving system in the Demo 2000 consists of 5 automated vehicles, each of which is equipped with the differential GPS (DGPS) for the localization, the laser radar for the obstacle detection and the inter-vehicle distance measurement, the inter-vehicle communication unit, and an onboard display unit. The ground GPS station for the DGPS has been installed on a test track, and the precision of the localization is within 2 cm under a still condition of a vehicle on the test track. The onboard display is for the passengers, and shows the results of the communications including the locations of the neighboring vehicles, their maneuver and intention. Fig. 3 shows the configuration of the vehicle. In addition to the units, each vehicle has the sensors for the speed, acceleration, yaw rate, and engine revolutions for the data logging and vehicle control. A 1 the vehicles can autonomously drive, and do not require 1 any infrastructure intelligence except for the GPS. The order of the vehicles was fixed in the demonstration, and only the lead vehicle has an additional function of the following vehicle detection by millimeter wave radars. The other 4 vehicles have the same performance. Fig. 4 shows the 5 vehicles. IV. VEHICLE CONTROL

neighboring vehicles does not always provide accurate data for the vehicle control, and sometimes the measurement becomes impossible. In addition, the transmission of emergency from a vehicle to the neighboring vehicles will require the inter-vehicle communications. The inter-vehicle communications is, thus, an essential technology to the cooperative driving.

The automated vehicles in the demonstration have automated lateral and longitudinal control functions.
A . Lateral Control Algorithm

The lateral control for a free agent or a lead vehicle in a platoon is based on the dead reckoning function with the DGPS. When the vehicle is steered with the dead reckoning

Fig. 3. Configuration of the automated vehicle.

Fig. 4. The automated vehicles for the Demo 2000.



Fig. I . Derivation of the lateral control algorithm.

function, it uses a precise map that contains a series of points VMIUE representing a course of the vehicle. An automated vehicle is steered to hit the point one after the other. Let the location of the current point that the vehicle is supposed to hit be ( x , ,y , ) , Fig. 6. The longitudinal control scheme. and the heading of the vehicle there be 8, on the vehicle coordinate. system as shown in Fig. 5. The current lateral from the brake table, and if it is positive, the engine torque is control 6 that drives the vehicle to hit the point with the calculated to find the throttle opening in the torque map heading is given as: together with the engine revolutions. 6 = arctan[2~(3y,- xl tanel)/ x: 1 (1) V. INTER-VEHICLE COMMUNICATIONS \viler: 1. is the wheelbcse of the vehicle [12]. Replacement of The inter-vehicle communication for the demoxtration was the point on: after the other whenever the vehicle approaches it will drive thc vehicle to the goal. This algorithm is also based on the 5.8 GHz DSRC, and it was designed for the: :ipplied to fol:owing vehicles in a platoon. In this case, the dedicated use in the demonstration. location of the preceding vehicle as well as the points on the A . Kequirementsfor the Inter-vehicle Conimunicarions map can be the point to be chased and hit. The requirzments for the inter-vehicle cormnunizations in B. Longitudinal Control '4 lgorithm the cooperative driving applications are the real time data The longirudinal control of a free agent or a lead vehicle is transmission characteristic that is essential to the vehicle tht speed control. The speeds, the acceleration, and the cor.trol, and the flexible networking Characteristic that is deceleration of the vehicle are programmed in advance taking essential to the changes in the topology of the nehvork and rhl: account of the speed limits, the maximum acceleration and number of the stations (vehicles), which are conflicting. If the deceleration, and the road geometry in the demonstration. stress is put on the real time data transmission characteristic, They are found in a table with the location of the vehicle the inter-vehicle communication protocol must be based on th: identified by the DGPS. token passing, and if the stress is put on the flexible The longitudinal control of the following vehicles is found networking characteristic, the protocol must be based on the with the inter-vehicle distance measured by the laser radar and CSMA (Camer Sense Multiple Access). Because in the calculated from thz localization data transmitted from the cooperative driving the network changes whenever a new preceding vehicle over the inter-vehicle communications. In vehicle joins the network or a vehicle leaves the network, by the demonstration, the speed command to the following putting the stress on the flexible networking characteristic, the CSMA has been employed in this study. vehicle v, is given as follows:
V'. = v p + k l ( v ; ,- V f ) + k , ( L r - L , , ) wliese kl =mlJL,-~,,,J/JL,J,kz = m * k , I)? : the current speed of the preceding vehicle,

,, ,




v, : the current

speed of the following vehicle,

L, : the reference inter-vehicle distance,
L,, : the measured inter-vehicle distance, and

m,,m, : control gains. Fig. 6 shows the longitudinal control scheme with the input of the speed command vc . After the acceleration is calculated with v, and v J , if it is negative, the braking control is found

B. Pufonnance o CSMA f The CSMA is classified into two classes: the non-persistent CSMA and the p-persistent CSMA. The perfoimance of each class has been evaluated with simulation studies [lo], and the results regarding the packet loss show that the non-persistent CSMA has better performance than the p-persistent CSMA does, when the transmission rate is below 1.28 Mbps, and the slot length is between 20 msec and 40 msec, the conditions of which are consistent with the applications of the inter-vehkle communications to the cooperative driving. From the MAC (Media Access Control) point of view, the delay in the processing is a main cause of the packet loss. The influence of the delay in the processing on the packet loss has been also investigated with the simulation studies. The


I EO + O

v1 c




a 1.E-03


I E-04






Success rate

Fig. 7. Mean packet loss rate vs. success rate.

non-persistent CSMA also shows better performance than the p-persistent CSMA does, when the delay is within 100p sec , thc transmission rate is 512 kbps, and the communication period is 20 msec. The non-persistent CSMA, thus, has been employed for the inter-vehicle communications protocol in the demonstration. The protocol has been nicknamed Dolphin (Dedicated Omni-purpose inter-vehicle communication Linkage Protocol for HIghway automatioN), following the ecology of the dolphins.

C. Implementation The MAC layer protocol data unit in the protocol for the demonstration consists of the packet control field (1 byte), the vehicle identification field (I byte), and the data field (30 bytes). The data field includes the vehicle location (6 bytes), the vehicle speed (2 bytes), the vehicle heading (3 bytes), and the obstacle location (4 bytes). The communication period among the 5 vehicles is 20 msec, and that between the onboard control and the communication control unit in each vehicle is IO0 msec. The real time data transmission performance of the inter-vehicle communications implemented was measured with the 5 automated vehicles on the test track. Fig. 7 shows the results. While one set of data is transmitted between the onboard control and the communication control unit, the inter-vehicle communications are performed 5 times. The collision of the packet causes the packet loss. When the the requirement on the packet loss rate is to be below requirement of the packet loss is satisfied 2 times out of 5 times of the communications among the parked vehicles, and 1 time out of 5 times among the driving vehicles. This is the reason why the transmission rate between the onboard control and the communication control unit is 100 msec. The protocol, thus, has the real time data transmission characteristic with the allowance of 100 msec. VI. DEMO 2000 SCENARIO

Fig. 8. Merging: 2 platoons are driving side by side (top) and one platoon merges into the other alternately (center, bottom).

the oval-shaped test track of 3.2 km long of Mechanical Engineering Laboratory (currently National Institute of Advanced Industrial Science and Technology) in Tsukuba at the end of November, 2000. The events in the scenario that was played out while making three circuits of the test track included:

In order to show the feasibility and potential of the
technologies, the demonstration “Demo 2000 Cooperative Driving” was held with a platoon of 5 automated vehicles on


stop & go of the platoon, platooning, splitting into two platoons assuming at an exit ramp,






F ,



Fig. 9. The preceding vehicles from the last one just before the obstacle detection.





Fig. i o . T h e onbosrd display on the event of the obstacle avoidance.




merging into one platoon from two platoons like at a ramp as shown in Fig. 8, passing by the last vehicle: the vehicles except the last one change lanes and the last passes the 4 vehicles like an ambulance, and then, the ambulance stops, and the 4 vehicles changes lanes to allow the ambulance to join the 4 vehicle platoon, obstacle detection and avoidance by lane changing: after one platoon splits into two platoons, the lead vehicle of a platoon detects an obstacle, and the platoon changes lanes and merges into the other platoon. Fig. 9 shows the preceding vehicles from the last vehicle (vehicle 5 ) under the cooperative driving just before the obstacle detection, and Fig. IO shows the onboard display on the obstacle avoidance, and platoon leaving and platoon joining.










Record cycle (2OOmsec)

Fig. 11. Vehicle speeds.

Most of the events were initiated by the location of the lead vehicle, but the merging on the obstacle detection was triggered by the detection of the obstacle (a parked vehicle) by the laser radar. The passing and platoon leaving were started when the driver operated the turn signal and the hazard lamps, which changed the mode of the vehicle from automated driving to manual driving.

Fig. 11 shows the speeds of the vehicles during the demonstration driving according to the scenario. As indicated in the figures, the vehicles drove at 40 - 60 k d h . The longer the distance between the lead vehicle and a following vehicle becomes, the worse the longitudinal control performance becomes. It is because the longitudinal control of each vehicle depends only on the preceding vehicle. Fig. 12 shows the inter-vehicle distances during the demonstration, which were calculated with the localization data acquired in each vehicle. The reference distance was 20 m in a platoon on a single lane roadway, and 34 m in platooris on a two lane roadway. At some parts, the inter-vehicle distances are larger than the reference distance. It is because the GPS signals could not be received at some parts of the test track, and the longitudinal control actuator was not sufficiently precise.



120 80

investigation on the increase in the throughput by the cooperative driving, and the deployment scenario of the systems equipped with the inter-vehicle communications. ACKNOWLEDGMENT A part of this work has been conducted under the cooperative research with Mechanical Engineering Laboratory and Association of Electronic Technology for Automotive Traffic and Driving, and the work of which been entrusted from Mechanical Social Systems Foundation. REFERENCES



2 -40


1 Change to following target vehicle (vehicle I)]







I Failme of DGPS and acceleration (aive by laser mdar) 1

[3] [4]
















and slow down like an ambu* [XI [9]










Record cycle (200msec)

[I I]

Fig. 12. Inter-vehicle distances.


K. S . Chang, J. K. Hedrick, W.B. Zhang, P. Varaiya, M. Tomizuka, and S . Shladover, “Automated Highway System Experiments in the PATH Program,” IVHSJournal, vol.1, no.1, pp.63-87, 1993. S . Tsugawa, “Vision-Based Vehicles in Japan: Machine Vision Systems and Driving Control Systems,” IEEE Trans. Industrial Electronics, vo1.41, no.4,pp.398-405, 1994. B. Ulmer, “VITA 11 - Active Collision Avoidance in Real Time,” Proc. IEEE Intelligent Vehicles 1994 Symposium, pp.1-6, 1994. C. Thorpe, T. Jochem, and D. Pomerleau, “The 1997 Autonated Highway Free Agent Demonstration,” Proc. 1997 IEEE Conference on ITS, pp.496-501, 1997. 0. Gehring, and H. Fritz, “Practical Results of a Longitudinal Control Concept for Truck Platooning with vehicle to Vehicle Communicaticn,” Proc. 1997 IEEE Conference on ITS, pp.117-122, 1997. M. Bertozzi, A. Broggi, and A. Fascioli, “Vision-based intzllipent vehicles: State of the art and perspectives,’’ Robotic and Aufonomous ~ysfems,vo1.32,no.l,pp.l-16, 2000. S. Tsugawa, “Super Smart Vehicle System: Future Intelligent Driding and the Measures for the Materialization,” Proc. IVHS Americs 1993 AnnualMeeting, pp.192-198, 1993. H. Fujii, 0. Hayashi, and N. Nakagata, “Sxperimental Research on Inter-Vehicle Communication using Infrared Rays,” Proc. IELE Inrelligent Vehicles I996 Symposium, pp.266-27 1, 1996. S . Kato, and S . Tsugawa, “Cooperative Driving of Autmomous Vehicles Based on Precise Localization with DGPS and Inter-Vehicle Communications,” Proc. 5th International Symposium on Advanced Vehicle Control, pp.261-268, 2000. K. Tokuda, ‘‘Inter-Vehicle Communications Technologies for Demo-2000,” Proc. IEEE Intelligent Vehicles 2001 Symposium, pp.339-344, 2001. S . Tsugawa, S . kato, K. Tokuda, T. Matsui, and H. Fujii, “An Overview on Demo 2000 Cooperative Driving,” Proc. IEEE Intelligent Vehicles 2001 Svmposium, pp.327-332,200 I . S. Tsugawa, and S. Murata, “Steering Control Algorithm for Autonomous Vehicles,” ISCIE/ASME Proc. 1990 Japaji - U S. A . Svmposium on Flexible Automation, pp.143-146, 1990.



Thc demonstretion on the cooperative driving has shown the feasibility and potential of the technologies of the inter-vehicle communications and the automated vehicle control. lmprovements on the inter-vehicle communications and the longitudinal control will be necessary. The inter-vehicle communications for the vehicle control should not have any delay. Since the occasional data loss can be compensated by estimation and prediction with the Kalman filtering technique, the protocol must be designed with the assumption that the occasional data loss can be allowed and the continuous data loss cannot be allowed. Precise longitudinal control would improve the performance of the formation of the cooperative driving. In addition, the hture issues may also include an algorithm for autonomous formation of the cooperative driving, the


To top