基于深度强化学习的固体放射性废物抓取方法研究

针对固体放射性废物分拣作业中，放射性废物杂乱无序、远程遥操作抓取效率低、人工分拣危险性大等典型问题，提出一种基于深度强化学习的放射性固体废物抓取方法。该方法使用改进深度Q网络算法，通过获取的图像信息，使机器人与环境不断进行交互并获得回报奖励，回报奖励由机械臂动作执行结果和放射性区域内放射性活度的高低构成，根据◢Q◣值的大小得到机械臂的最佳抓取位置。用V-REP软件对UR5机械臂建立仿真模型，在仿真环境中完成不同类型固体放射性废物抓取的训练与测试。仿真结果表明，固体废物在松散放置时该方法可使机械臂抓取成功率大于90%，在紧密放置时抓取成功率大于65%，机械臂不会受到废物堆叠的影响，并且会优先抓取放射性区域内具有高放射性活度的物体。

Research on a solid radioactive waste grasping method based on deep reinforcement learning

In the solid radioactive waste sorting operation, in order to solve the typical problems of disordered radioactive waste, low efficiency of remote teleoperation, and high risk of manual sorting, this paper proposed a method for grasping radioactive solid waste based on deep reinforcement learning. This method used an improved depth Q network algorithm. Through the acquired image information, the robot could continuously interact with the environment and obtain rewards, which were composed of the execution results of the robotic arm actions and the size of the radioactive activity in the radioactive area. According to the size of the Q value, the robot arm got the optimal grasping position. V-REP software established the simulation model of the UR5 robot arm, and different types of solid radioactive waste grasp training and testing were completed in the simulation environment. The simulation results show that when the radioactive solid waste is loosely placed, the method can make the grasping success rate greater than 90%, and the grasping success rate is more than 65% when the radioactive solid waste is placed tightly. The robotic arm grasping operation is not affected by the stacking of objects and the objects with high radioactivity in the radioactive area are preferentially grasped.