基于力觉引导的机械臂自适应开门旋拧方法

针对核应急环境中,环境模型未知、人工开门危险性较大的问题,提出了一种基于力觉引导的机械臂自适应开门旋拧方法.该方法通过机械臂末端的六维力传感器获得力和力矩信息,将实际力或力矩与期望力或力矩之间的差值作为深度确定性策略梯度算法的状态输入,同时输出动作;利用机械臂末端所受两个方向力的函数关系,设置基础奖励函数,通过机械臂的期望运动方向,设置引导性奖励函数,使机械臂自动适应力与力矩的变化,完成旋拧门把手任务.仿真数据结果表明,在有引导性奖励的情况下,基于力觉引导的机械臂自适应旋拧方法能够在更短的时间内达到收敛,完成机械臂旋拧门把手的任务.

Adaptive opening and screwing method of manipulator based on force guidance

In the uncertain nuclear environment, the manual door opening is dangerous. This paper proposed an adaptive opening and screwing method of manipulator based on force sense guidance. The six-dimensional force sensor could obtain the force and moment information at the end of the manipulator. It input the difference between the actual force or torque and the expected force or torque into deep deterministic policy gradient algorithm, output actions at the same time. This method used the function relationship between the two direction forces at the end of the manipulator to set the basic reward function and used the expected movement direction of the manipulator to set the guided reward function, for the manipulator could automatically adapt to the change of the moment and the stress, and completed the task of screwing the door handle. The experimental results show that the self-adaptive rotation method based on force guidance can achieve convergence in a shorter time and complete the task of rotating the door handle.