1. 1 Utilizing Shared Vehicle Trajectories for Data Forwarding in Vehicular Networks IEEE INFOCOM MINI-CONFERENCE Fulong Xu, Shuo Gu, Jaehoon Jeong, Yu Gu, Qing Cao, Ming Liu and Tian He Computer Science and Engineering University of Minnesota April 11rd, 2011

2. Motivation 2  The vehicular networking is getting a hot research topic.  Internet Access, Driving Safety, Data Dissemination, etc.  The existing data forwarding protocols in VANETs  Many ones only take advantage of the road network layout and traffic statistics.  A few adopt available vehicle trajectories along with road traffic. (use the trajectory in a privacy-preserving way)  The objective in this paper  Utilizing shared trajectories to provide effective vehicle-to- vehicle (V2V) communications over multihops in VANETs.

3. Problem Formulation  The assumptions for the vehicular networks  Every vehicle has a GPS-based navigation system. Traffic statistics are available via commercial navigation services.  The V2V communication operates in a participatory manner. To obtain the communication service, a vehicle should shares its trajectory with other participated ones.  The Access Points (APs) are sparsely deployed in road networks. They are interconnected and disseminate vehicles’ real time trajectory information.

4. Basic Idea 4 If the vehicle Va want to send data to the Vc  STDFS is based on vehicular encounter prediction Packets can be forwarded through the “encountered vehicles path”: Va → Vb → Vc.

5. Contribution and Challenges 5  Contribution  Data forwarding based on Shared Vehicle Trajectory With shared vehicle trajectory, STDFS outperforms the existing scheme (VADD and TBD).  Challenges  Pair-wise encounter prediction and the construction of a vehicle encounter graph Mathematical model for the travel time  Optimization of the encounter graph to achieve a low delivery delay under the required delivery ratio threshold

6. Travel Time Prediction 6  Basic Theory  The travel time of one vehicle over a fixed distance follows the Gamma distribution.  Therefore, the travel time through a travel path in the road network is modeled as: and can be calculated using the traffic statistics.

7. Pair-Wise Enconter Prediction  The probability of V a encounter V b at road section L 12 is:  The “ 12 ” means “encountering at road section L 12 ”.  Its transformation is:  T a1 and T b1 are independent stochastic variables following gamma distribution.

8. Conditional Encounter Probability Calculation in Multi- hop Encounter Prediction  Why use the conditional encounter probability?  It is used for multi-hop encounter prediction.  If Va want to send packets to Vc,  The success probability is:  An approximate method is used to calculate this conditional probability. Conditional Encounter Probability

9.  Why to calculate the expectation of two vehicle’s encounter time?  It is used in the process of constructing the encounter graph.  Let the encounter time is T, the expectation of the encounter time is.  T is a function of T a1 and T b1.. The expectation of two vehicles’ encounter time encounter position

10. Constructing a Predicted Encounter Graph  The predicted encounter graph is a directed graph.  originates from the source vehicle that intends to forward packets  ends at the forwarding destination  For a node e in the Graph,  Its child nodes are the vehicles it might encounter later after its parent;  Its child nodes are sorted in the sequence of their expected encounter time with node e.

11. Constructing a Predicted Encounter Graph  The construction is a process of expanding the graph by adding new nodes one by one, according to the sequence of the expected encounter time. Queue: Graph:

12. Three Definitions 1.Forwarding Sequence  includes n vehicles (children) that can forward packets from vehicle e to the destination.  This sequence is sorted by the expected encounter time with its parent (vehicle e).

13. Three Definitions 2. Expected Delivery Ratio (EDR):  The expected delivery ratio of a given vehicle e, denoted by EDR e, is the expected packet delivery ratio from vehicle e to its destination.  If vehicle e’s i th forwarder’s EDR value is EDR i, Forwarding Sequence EDR 1 = 80%EDR 2 = 60%EDR 3 = 70% 0.4 0.9 0.2 EDR e =0.4*0.8 + (1-0.4)*0.9*0.6 + (1-0.4)*(1-0.9)*0.2*0.7

14. Three Definitions 3. Expected Delivery Delay (EDD):  The expected delivery delay of a given vehicle e, denoted by EDD e, is the expected data delivery delay for the packets sent by vehicle e and received by the destination.  If vehicle e’s i th forwarder’s EDD value is EDD i,

15. Optimizing Expected Delivery Ratio (EDR)  For vehicle e’s full forwarding sequence  So we should only choose a subset of the forwarders and get the optimal forwarding sequence! 1.0 EDR 1 =0.6 EDR 2 =0.9 Should all the forwarders in the sequence be selected to forward packets?  If both node 1 and node 2 are selected as forwarding nodes: EDR e = 1*0.6 =0.6  If only node 2 are selected as forwarding nodes: EDR e = 1*0.9 =0.9

16. Optimizing Expected Delivery Ratio (EDR)  Select only a subset of the forwarders  node 3 has to be selected for data forwarding.  Then try to add more nodes into the optimal forwarding sequence backwardly. EDR 1 = 80%EDR 2 = 60%EDR 3 = 70% 0.4 0.9 0.2

17. Optimizing Expected Delivery Delay (EDD)  It is meaningless if only optimize EDD and do not care about EDR.  Our goal is to optimize the EDD metric for the root node under the constraint that the EDR metric is no less than a certain threshold R.  Since the predicted encounter graph is expanded in the order of expected encounter time, it is helpful to optimize EDD.

18. Optimizing Expected Delivery Delay (EDD)  The method to optimize EDD In the process of constructing the graph, when a new path to the target node is found, use the approach of optimizing EDR to calculate the EDR of the root node:  If the the EDR value is greater than the required bound R, the graph construction stops and the optimal forwarding sequence is acquired.  Otherwise the expanding continues.

19. Forwarding Protocol 19  STDFS Forwarding Rule  Within a connected component, packets are forwarded to the best forwarder.

20. How to select the best forwarder?  Within the connected component, each vehicle calculates its own EDR and EDD:  If the EDRs of all the connected vehicles can not meet the requested bound R, the vehicle having the highest EDR is the best forwarder;  If there exists the vehicles whose EDRs are greater than the bound R, within these vehicles the one having the minimal EDD value is the best forwarder.

21. Performance Evaluation 21  Evaluation Setting  Performance Metric: (i) Data Delivery Ratio, (ii) Average Delivery Delay  Parameters: (i) Vehicle speed deviation, (ii) Vehicular traffic density. *We focus on data forwarding from vehicles to fixed points.  Simulation Environments  36-intersection road network (4.2 miles X 3.7 miles)  Vehicle mobility model: Manhattan Mobility model  Vehicle speed distribution: N(40,7) MPH  Communication range: 200 meters  Time-To-Live (TTL): 1000 seconds  Requested EDR bound R : 0.9

22. Impact of Vehicle Speed Deviation STDFS outperforms VADD and TBD under different vehicle speed deviations. As the vehicle speed deviation increases, STDFS has a higher delivery delay.

23. Impact of Vehicular Density STDFS is more suitable for data forwarding when vehicular networks become sparse.

24. Conclusion 24  In this talk, the data forwarding scheme called STDFS is introduced based on the vehicle trajectory:  Data Forwarding from Vehicle to Vehicle.  Also, the predicted encounter graph is introduced for STDFS data forwarding scheme:  This predicted encounter graph can be used for other VANET routing or forwarding schemes.  As future work, the privacy issue caused by sharing trajectories with public will be studied.