Collaborator: A Nonholonomic Multiagent Team for Tasks in a Dynamic Environment

In our previous work, we proposed a potential field based hybrid path planning scheme for robot navigation that achieves complete coverage in various tasks. This paper is an extension of this work producing a multi-agent framework, Collaborator, that integrates a high-level negotiation-based task allocation protocol with a low-level path planning method taking into consideration several real-world robot limitations such as nonholonomic constraints. Specifically, the proposed framework focuses on a class of complex motion planning problems in which robots need to cover the whole workspace, coordinate the accomplishment of a task, and dynamically change their roles to best fit the task. Applications in this class of problems include bomb detection and removal as well as rescuing of survivors from accidents or disasters. We have tested the framework in simulations of several tasks and have shown that Collaborator can satisfy nonholonomic constraints, cooperatively accomplish given tasks in an initially unknown dynamic environment while avoiding collision with other team members. Finally we prove that the proposed control laws are stable using the Lyapunov stability theory.