中文摘要：

为了解救陷入环境障碍的自主移动机器人,提出了一种基于强化学习的自救脱困控制方法.该方法通过移动机器人与环境的交互作用,能够在线学习实现脱困自救的运动控制策略,并利用机器人自身条件克服环境障碍,避免了实施救援机器人的行动和终止其作业任务所造成的损失.利用工作环境的先验知识指导,设计含有启发信息的强化学习系统回报函数,保证搜索和学习控制策略向正确方向进行,同时提高学习控制器的适应性和鲁棒性.数字仿真证明了通过自学习控制策略实现自救脱困的可行性.

英文摘要：

A control technique to achieve self-rescue of autonomous mobile robot from obstacle environment based on reinforcement learning was proposed. Motion control strategy of self-rescue was got on line through the interaction between the mobile robot and the obstacle environment, so the self-rescue control technique helps to overcome obstacle environment by the autonomous mobile robot independently and avoid the loss from rescue activity and task failure. Prior knowledge of working environment was applied to direct the design of heuristic reward function for the reinforcement learning system, which guarantees the correct direction of searching and learning control strategy. The simulation experiments indicate that it is feasible to achieve self-rescue by self learning control strategy.