Adversarial blocking techniques for autonomous surface vehicles using model-predictive motion goal computation

In this paper, we consider the problem of guarding a valuable naval asset from a highly maneuverable threat via the use of autonomous unmanned surface vehicles (USVs) as dynamic obstacles. The objective of the defending agent is to maximize the amount of time it takes an intruder boat to enter the restricted area. Here we introduce a set of active blocking strategies which allow the defender to influence the intruder's actions through exploitation of its obstacle avoidance. By applying a mirror transformation to the intruder's current state, a pursuit policy can be formed that always places the defender between the intruder and the asset. To apply these strategies to delayed or imperfect information environments, we present a model-predictive method of calculating defender motion goals and estimating future intruder motions. This predictive estimator is capable of generating a meta-model for the intruder's behavior and returning a probability distribution of intruder control actions based on a reduced set of observable spatial features.