Unpredictable Trajectory Design for Mobile Agents

Mobile agents have attracted considerable attentions for their wide applications in civilian and military fields, where motion planning plays an important role when the agents operate in physical world. During this process, the agents are prone to path information leakage and malicious attacks on trajectories, which lead to individual malfunction or even mission failure. In order to protect the future position information contained in history trajectory and evade physical interception attacks, this paper studies unpredictable trajectory design for mobile agents. The major challenges lie in two parts. First, how to determine the optimal form of control for one agent, in face of unknown observation accuracy and prediction algorithm of the attacker. Second, how to extend the control method of one agent to multiple agents with coupled dynamics. The novelty of our work is threefold: i) Leveraging the stochastic control method, the trajectory design problem is formulated as optimization problems universal for various prediction methods; ii) In the sense of expectation and probability measure, we propose two kinds of optimization objectives which are considered synthetically, and obtain the optimal control for secure movement. iii) We extend the method to multiple agents in formation, and achieve a trade-off between the degradation of formation convergence and the improvement of safety level. Simulations demonstrate and verify the effectiveness of the proposed approach.