Effective Team Coordination Through Intra-Robot Replanning to Restore Team Plan Rationale

Abstract

We consider multiple autonomous underwater vehicles (AUVs) engaged in oceanic missions that require collaboration. In this paper, we address the challenge of effective distributed multi-robot plan execution: when a failure occurs, an individual robot needs to decide how to proceed to maintain the success of the multi-robot plan. Failures occur when the modeled actions of the AUVs fail to result in their expected outcome, largely due in this domain to the poorly modeled dynamics of the ocean. Intra-robot replanning is advantageous because traditional purely centralized controlled methods use time-consuming and expensive satellite communication in order to repair all failures. Therefore, we present a rationale-driven team plan representation that explicitly includes the rationale of the individual robots' planned actions. We then present our intra-robot rationale-driven algorithm that enables the selection and execution of replan policies for individual robots to overcome failure. This paper is focused on examples of concrete team missions and the collaboration of AUVs that clearly demonstrates the effectiveness of our approach.