Urban Vehicular Mobility Patterns For Driving Route Prediction

The development of wireless communication technology has brought profound changes to our work and life. As an important branch of wireless communication technology, vehicular ad hoc network will provide us with technical support for achieving the safe, comfortable, intelligent urban traffic environment.The complexity of urban traffic environment makes the research on vehicular ad hoc network quite challenge. Traditional wireless network simulation and verification methods can not be applied to vehicular network directly. How to build the vehicular network environment close to the real traffic environment is an important research field. This paper will focus on the characteristics of vehicular network and their impact on the network performance, based on the 4,000 taxis'real traffic data collected by Shanghai Grid project. In this paper, we mainly consider the following problems.First, construct the experiment platform for vehicular network research. Integrated and accurate data is the basis of our experiments. However, the traffic data collected by the GPS devices is discrete and inaccurate. This data can not be directly used for experimental analysis and some pre-work need to be done. We analyze the abnormal reasons for the GPS data error, match the vehicle location information to the road segments, select the most suitable driving route from potential routs, and interpolate new data to reconstruct the vehicles driving trajectories.Second, analyze the experimental data and extract vehicle mobility patters. As we know, human's social habits have certain spatial and temporal regularity. Through analyzing large amounts of vehicle traffic data, we found that vehicle motion has some similar regularity. This regularity makes vehicles have some mobility patterns on some roads. To extract these mobility patterns, we apply variable-length Markov model to train and generate the mobility patterns. Compared with traditional Markov models, variable-length Markov model can easily extract vehicle mobility patterns with different orders.Third, predict the vehicle driving route. Vehicle mobility patterns can be used for short-term route prediction. If a vehicle has mobility patterns on some road, based on the roads the vehicle just passed, we can predict the vehicle upcoming route with a high probability. According to the experimental results, we know that route prediction based on the vehicle mobility patterns can obtain high performance.Fourth, analyze the traffic conditions'impact on vehicle mobility patterns. Traffic conditions are dynamic changing according to time. To assess its impact on the vehicle mobility patterns, we consider the traffic conditions as an important argument during the vehicle mobility pattern extraction and route prediction. Experimental results show that traffic conditions can influence the vehicle driving route selection and mobility patterns.Fifth, apply the vehicle driving route prediction in the vehicular network. The motion characteristics of vehicular network have an important impact on the network performance. Taking advantage of the vehicle driving route prediction, we can optimize the design of network protocols. In our work, we utilize the vehicle driving route prediction during the routing protocol design and data transmission mechanism design. Experimental results verify that the vehicle driving route prediction can help to improve the network performance.