Multi-Vehicle Tracking and State Estimation through Data Association 2020-01-5149
Internet of Things (IoT) devices are playing increasingly important roles in road sensing of Intelligent Transportation Systems (ITS). For instance, through leveraging small sensors on roadside, the volumes of driving vehicles and vehicle driving speeds are obtained, which are essential to road traffic management. However, clutter data issue in high density road traffic and multiple sensors synchronization problem can reduce possibilities of successful vehicle detection. In this paper, we present a novel traffic volume counting and vehicle state estimation method using multiple on-road sensors. Taking account for the clutter in dense environment and multi-sensor registration problems, we categorize vehicle tracking and state estimation methods as vehicle data association problem. A multi-vehicle tracking model is formulated to monitor the road traffic condition, which is capable of tackling with the issues of missing data issues, multi-sensor time synchronization, detected vehicle number and traffic flow speed. To efficiently solve the clutter problem, a probabilistic density analysis is developed. An algorithm is proposed based on multi-sensor multi-vehicle architecture. Moreover, a method for time synchronization is proposed to analyze vehicle tracking. The validation region in each sensor is adapted to the current measurement settings, which are unified in the optimal estimates. Through modeling measurements of sensors in non-uniform estimation rate, we unify the time synchronization for multi-sensor fusion. The proposed vehicle tracking and estimation technique are validated with Monte Carlo simulation and on-road experiment data. On the basis of the experimental estimations for each validated measurement corresponding to vehicle tracks in clutter environment, it is concluded that our proposed solutions can effectively estimate vehicle states and increase accuracy of tracking performance.