Cooperative Motion and Task Planning Under Temporal Tasks

Abstract: Temporal-logic-based languages provide a formal and accurate way to specify complex motion and action missions for autonomous robots, beyond the classic point-to-point navigation task. The first part of the thesis is devoted to the nominal scenario: an autonomous robot is given a motion task specified as Linear-time Temporal Logic (LTL) formulas. Under the assumption that the workspace is static and fully-known, we provide a systematic and automated scheme to synthesize both the discrete motion and task plan and the hybrid control strategy that drives the robot, such that the resulting trajectory fulfills the given task specification. Limited knowledge about the workspace model, unforeseen changes in the workspace property and un-modeled dynamical constraints of the robot may render the nominal approach inadequate. Thus in the second part of the thesis we take into account four non-nominal scenarios where: (i) the specified task is not feasible; (ii) the task contains hard and soft constraints; (iii) the workspace model is not fully-known in priori; (iv) the task involves not only robot motion but also actions. The proposed results greatly improve the real-time adaptability and reconfigurability of the nominal scheme. In the last part, we analyze a team of interconnected autonomous robots with local and independently-assigned tasks. Firstly we consider the case where cooperations among the robots are imposed due to heterogeneity and collaborative tasks. A decentralized coordination scheme is proposed such that the robots' joined plans satisfy their mutual tasks the most. Then a distributed knowledge transfer and update procedure is designed for the networked robots that co-exist within a common but partially-known workspace. It guarantees both the safety and correctness of their individual plans.