Data Communication And Traffic Sensing In Vehicle Ad Hoc Networks

In the recent years, the number of vehicles in the cities is boosting,which causes a heavy burden for the public traffic system. Someresearchers and engineers are trying to approach this problem by buildingan Intelligent Transportation System (ITS). Vehicle Ad Hoc Networks(VANETs) are one of the most important components in the ITS. More andmore researchers and engineers focus on VANETs.VANETs connect the mobile vehicles of the city into an ad hocnetwork using some short-distance wireless technologies like DSRC andWi-Fi. So the vehicles can communicate with each other or with theroadside infrastructures. Based on VANETs, we can build a Probe VehicleSystem to collect traffic information it the city, which is much moreeffective and cheaper than traditional methods. However, there are stillsome great challenges now: i) The delivery rate in VANETs is low and thedelay is large; ii) There is no guarantees that traffic information can becollected in a certain location at a specific moment.In this paper, we propose a location based routing algorithm for thedata communication in VANETs. First, we introduce a practical locationservice, which can locate a vehicle in the networks for us. Then we bringin the carry-forward strategy in Delay Tolerant Networks. Specificallyspeaking, a vehicle will carry a data packet and move. When it encounterswith other vehicles, it will choose the best neighbor to relay based on ourgreedy strategy. We do some simulation experiments based on real GPStrace data and find that our algorithm is effective.Then, we extract the traffic condition information of the roads fromsome GPS trace data. Principle component analysis is used to analyzed these data. And we figure out that there is the hidden structure andredundancy in these sensing data. We try to utilize this nature feature toestimate the missing value caused by non-uniform sample problem. First,we define a matrix to denote the traffic condition for all roads in the city.The element of the matrix is the average driving speed. So a certain rowdescribes a time series of one road, while a certain column describes asnapshot of all roads in a given time slot. Because there is no a guaranteethat we can have a probe vehicle pass by and sample in a given time and agiven road, some elements in the matrix are missing. We propose acompressive sensing based recovery algorithm to estimate these elements.In our experiment, our algorithm can estimate the traffic condition matrixvery well using only a few data.