High-level task planning under adversarial environments is one of the central problems in the development of autonomous systems such as unmanned ground vehicles (UGV). Existing works commonly assume that the decision-maker such as UAV shares the same information with the environment. However, in many scenarios, the UGV, as an integral part of the system, generally has more information than the external adversary. For such a scenario, the decision-maker with more information may achieve better performance by using deceptive strategies. In this article, we investigate the problem of optimal deceptive strategy synthesis for autonomous systems under asymmetric information between the internal decision-maker and the external adversary. Specifically, we model the dynamic system as a weighted two-player graph game and the objective is to optimize the mean payoff value per task. To capture the asymmetric information between two parties, we assume that the UGV has complete knowledge of the system, whereas the adversary may have misconceptions regarding the task as well as the cost. To synthesize an optimal deceptive strategy, we propose a synthesis algorithm based on hyper-games. The correctness as well as the complexity of the algorithm are analyzed. We illustrate the proposed algorithm by running examples as well as a simulation case study. Finally, we conduct an empirical experiment using real-world scenarios to verify the practical applicability of our algorithm.
Optimal Deceptive Strategy Synthesis for Autonomous Systems Under Asymmetric Information
IEEE Transactions on Intelligent Vehicles ; 9 , 10 ; 6108-6121
2024-10-01
1587524 byte
Article (Journal)
Electronic Resource
English
Deceptive robot motion: synthesis, analysis and experiments
British Library Online Contents | 2015
|Deceptive Actions and Robot Collision Avoidance
Springer Verlag | 2019
|Deceptive Communications - When in doubt, verify
Online Contents | 2008