Interception algorithm for autonomous vehicles with imperfect information

Abstract

Autonomous vehicles often operate in environments with imperfect information. This thesis addresses the case of a system of autonomous vehicles and sensors attempting to intercept a moving object of interest that arrives stochastically and moves stochastically after arrival. A sensor array is placed in the area of expected arrivals. As the object of interest moves across the sensor system, the system initially receives perfect information of the object's movements. After the object of interest leaves the sensor system, the algorithm uses statistical estimation techniques to develop confidence intervals about points of expected interception. The algorithm assigns the optimal, autonomous chase vehicle from a set of pre-positioned autonomous vehicles, develops movement commands for the assigned vehicle, and considers reassignment of chase vehicles as appropriate given the stochastic movements of the object of interest. Dynamic programming is employed to optimize system parameters, and the thesis considers a reformulation of the problem that uses dynamic programming as a structural model for the entire algorithm.