Intention Obfuscated Adversarial Deceptive Path Recommendation for UGV Patrol Maneuver

The problem of adversarial Unmanned ground vehicles (UGVs) patrol has received increasing attention in recent decade. UGV patrol maneuver may encounter one fullknowledge opponent in the adversarial setting, which makes it difficult to ensure safety. In this paper, we propose one framework based on generative adversarial networks (GAN) to generate reliable paths and recommend the top-k paths based on deception score by employing some deceptive tactics during the path evaluation process. In adversarial setting, our proposed framework provides the UGV with an adversarial deceptive path that helps it maneuver from local position to the desired position. We demonstrate the recommendation process of an adversarial deceptive path in support of a path planning application for the UGV patrol maneuver. Our experiments show the feasibility and usefulness of the recommended paths.