Connected Car Networking

Transcription

1 Connected Car Networking Teng Yang, Francis Wolff and Christos Papachristou Electrical Engineering and Computer Science Case Western Reserve University Cleveland, Ohio

2 Outline Motivation Connected Car Technology Background Advantages Approach Initial static route prediction Dynamic traffic behavior anticipation Path planning prediction Results 2

3 Background Navigation-based methods not based on connected cars Current GPS-based or satellite-based applications limitations Is location aware Based on street maps Cannot provide lookahead, anticipation or prediction Cannot handle accidents in the resolution of seconds There other methods assisting in highway driving Connect car technology is not yet a developed technology 3

4 Motivation Exploit the emerging connected car technology Intelligent Transport Systems Standards Such as the IEEE DSRC protocol (Dedicated Short Range Communications) [1] The U.S. Department of Transportation s Connected Vehicle Program [2] Need a system of new navigation algorithms Assist drivers finding near optimal routing in real-time Provide intelligent transportation solution based on connected cars Improve traffic, reduce accidents Reduce energy carbon footprint Not based on GPS, GPS could be used for enhancement 4

5 Connected Car Network Technology The DRSC technology is based on these characteristics Connectivity Vehicle to vehicle, V2V Vehicle to infrastructure, V2I Wireless real time information sharing, 5.9 GHz Band (not ) [1] V2V, Data collect from car sensors from neighboring cars Speed of car, acceleration, distance car to car. Detect red lights, braking, etc. V2I, traffic lights, lane markers, streetlights, signage and parking meters Real-time advisories about such things as road conditions, traffic congestion, accidents, construction zones and parking availability. 5

6 Advantages Objective of our method is to develop a system (Hardware/Software) to assist drivers in optimal routing in real traffic time less than a seconds Focus on suburban rush hour traffic in large metropolitan areas Can handle blockage information ahead of time Our system can take advantage of this to re-route, path planning Anticipatory lookahead of the mitigation effects and predictive path planning 6

7 Approach: Initialization Based on pre-estimation of traffic flow First we have a street map of the area Every street associated with a speed limit: static limits What is the speed limit, 25 mph, school districts during rush hour, pre-assigned Characterized by speed limits & main roads, intuitively Convert map into a graph Nodes, Arcs and tags represent intersections, street segments, traffic information After initialization, the tags change dynamically As the car drives And new traffic information is received 7

8 Approach: Path Planning Information without Traffic Assuming one car is driving from Point A to Point R, the path from A to R can be divided into five segments. Segment Distance (miles) Speed Limit (miles/hour) A - F F - N N - Q Segment Time (seconds) 30 s N 30 s Q 10 s R Q - R where for t = 0, initial segment time is A 20 s F t segment i, j = d segment i, j v segment i, j Example t A, F = d A, F v A, F = 0.2 m 35 mph =20 s 8

9 Street map: tagged attributes The routing & planning information is tagged to each segment, i.e. t segment Segment tags also store V2V and V2I information tag ts (30) 30 s Q 10 s R tag ts (30) N tag ts (30) 30 s F A 20 s tag ts (20) J 9

10 Approach: Dynamic real-time Dynamic runtime Check if car is at destination node Begin Collect V2V and V2I information Anticipatory lookahead Compute traffic prediction Path Planning Generate possible new routes Alert driver of any better route changes Destination? Collect V2V & V2I (Blockage events) Anticipatory lookahead Predict New route Update Driver Loop Done 10

11 Dynamic V2V velocity information For traffic flow of a queue of cars, space mean flow can be used instead of average velocity v i, j (t)= Segment at time, t N n=1 N 1 v car i, j (t) Number of cars and Cars with Diferent Speed (mile/hour) A - B 5 35, 30, 25, 35, 30 B - C 7 35, 30, 25, 35, 30, 35, 30 C - D 7 35, 30, 25, 35, 30, 30, 30 t segment i, j (t)= d i, j v i, j (t ) Estimated Time 23.6 s 34.8 s 35.5 s A 23.6 F 34.8 N 35.5 Q 13.2 R D - E 5 30, s 11

12 Blockage Events and Effects Multiple street Congestion result in possible blockage effects Need to propagate and predict new traffic conditions. Secondary roads can take priority over primary roads Updated Route map Congestion Congestion J J Congestion 12

13 Traffic anticipation and path plan prediction The first driver stops suddenly, The second driver sees red brake lights and now brakes. This continues until several cars later also come to a halt. The problem is that the driver wants to know is going to happen, instantaneously, several cars ahead, in order to make a decision Just knowing when the first car brakes does not tell the driver how the rest of the cars behave Want to anticipate and mitigate any possible action Show the Driver wait or take new route? Connect Cars will tell me Event propagation, blockage, physical process Data propagation, instantaneously from car to car, data process Need a Kinematic formula that predicts the delay 13

14 Traffic prediction in order to change route t break = v 1 a and t prediction = 2 S n = n t v reaction + a t 2 break 1 v 1 S n =n (d 1 d 2 )+S 1 =n S r +S b 14

15 Sample Results Situation 1 Best routes initially (Total 16s) A-B-C-D-E-G-M-P-R and A-F-N-Q-R Situation 2 Blockage event between A & B (+7s delay) A-B-C-D-E-G-M-P-R (23s) best route A-F-N-Q-R (16s) Situation 3 Accident event between Q & R (+2s delay) best route A-B-C-D-E-G-M-P-R Situation 4 multiple accidents N&Q, M&P, (+4,+2s) best route A-B-C-D-E-G-M-P-R (18s) Note: normalized time units 15

16 Prototype implementation System implementation Hardware Processor board Raspberry Pi Wireless communication module Software implementation DRSC, Dedicated short-range communications Path Planning algorithm Dashboard display 16

17 Conclusion Using Connect Car Technology, developed a Navigation system for software and hardware A system which produces in realtime optimal navigation routes in real suburban rush traffic Takes advantage of existing Connect Car Technology and DRSC protocol Realtime less than a second, route is updated in seconds, route updated all the time Traffic prediction goes beyond a simple reaction algorithm of event Lookahead anticipation predicts what other cars will behave before the V2V confirms it This makes for even more advanced path planning prediction Realtime traffic takes accounts of blockages, traffic lights Motivation came from the Department of Transportation s website The prediction algorithm takes kinematic and if there is a blockage, prediction traffic delays due to blockage 17

18 References 2008, Bob Williams, Intelligent Transport Systems Standards, Artech House, Technology & Engineering, 836 pages. The U.S. Department of Transportation s (USDOT s) Connected Vehicle Program, ITS Research CONNECTED VEHICLES, , Traffic and Highway Engineering, Nicholas J. Garber, Lester A. Hoel Cengage Learning, 1248 pages 18

19 Questions? 19