基于Kinect传感器的移动机器人环境检测及行为学习

Mobile robot environment detection and behavior learning based on Kinect sensor

研究了一种基于深度图像和强化学习算法的移动机器人导航行为学习方法。该方法利用机器人装配的Kinect传感器检测工作环境信息,然后对获取的深度图像数据和视频图像进行处理、融合和识别,并由此构建机器人任务学习的状态空间,最终利用强化学习方法实现移动机器人的导航任务的自学习。该方法的有效性通过实验得到验证。实验表明,该方法能够使机器入具有较强的环境感知能力,并能够通过自学习的方式掌握行为能力。

A behavior learning algorithm for mobile robot navigation based on depth images and reinforcement learning is proposed. The algorithm uses the Kinect sensor on a mobile robot to capture the environmental data of the robot, then, processes, fuses and identifies the data of depth images and video images among them to establish the state space for robot learning, and finally, uses the reinforcement learning method to implement the mobile robot＇ s self- learning of navigation tasks. The effectiveness of proposed algorithm was verified by experiment. The experimental results show that the method can make a mobile robot posses the stronger ability of perceiving environments and ca- pacity of mastering behaviors by self-learning.