Topics in high-level robot control: Integrating planning and reactivity, and multiple-robot coordination.

摘要

It is difficult to build an autonomous robot controller that can control a robot to perform complex tasks. It should not only generate plans for the robot to accomplish tasks, but also be able to handle unexpected events and run-time failures with incomplete knowledge of the environment. This thesis presents a robust approach for constructing high-level robot control programs using the agent-oriented programming language IndiGolog. Its predecessor ConGolog has been shown to be an effective programming language for constructing reactive and deliberative agent programs. By improving the planning and plan execution mechanisms in IndiGolog, planning and reactivity can be integrated effectively to optimize the actions performed by the robot while maintaining robustness in a dynamic environment. Our approach is also applied on a real shipment delivery application with multiple robots, to demonstrate how one can build a distributed control system to coordinate multiple robots to accomplish a task using IndiGolog.