基于ai深度学习的机器人碰撞预估计控制器设计
2023,31(5):160-165
摘要：机器人运动过程中与外部障碍物之间容易发生碰撞，当碰撞作用力过大时会造成机器零件损坏的问题，为解决这一问题，设计基于ai深度学习的机器人碰撞预估计控制器。建立人机交互电路与串口通信电路，将伺服电机设备、运动控制器、PC感应装置分别接入既定作用区域内，完成预估计控制器的整体应用结构设计。以PyTorch深度学习框架为基础，定义激活函数，再根据预估计参数的实际取值范围，实现对目标机器人对象的精准检测。按照力矩控制条件表达式，确定碰撞行为的表现强度，完成对机器人运动路径的规划，联合相关应用设备，实现基于ai深度学习的机器人碰撞预估计控制器设计。实验结果表明，ai深度学习算法作用下，机器人与障碍物碰撞部位的接触面积不会超过0.25m2，由碰撞行为导致的外部作用力相对较小，不会造成严重的机器零件损坏问题。
关键词：ai深度学习；机器人碰撞；预估计控制器；人机交互；PC感应装置；PyTorch框架；激活函数；
Design of robot collision pre estimation controller based on ai depth learning
Abstract：It is easy to collide with external obstacles in the process of robot movement. when the collision force is too large, it will cause damage to machine parts. to solve this problem, a robot collision pre estimation controller based on ai depth learning is designed. The human-computer interaction circuit and serial communication circuit are established. The servo motor equipment, motion controller and PC induction device are respectively connected to the given action area to complete the overall application structure design of the pre estimation controller. Based on the PyTorch depth learning framework, the activation function is defined, and then the accurate detection of the target robot object is realized according to the actual value range of the pre estimated parameters. According to the torque control condition expression, the performance strength of the collision behavior is determined, the robot motion path planning is completed, and the robot collision pre estimation controller design based on ai depth learning is realized by combining relevant application equipment. The experimental results show that under the action of ai depth learning algorithm, the contact area of the collision part between the robot and the obstacle will not exceed 0.25m2, and the external force caused by the collision behavior is relatively small, which will not cause serious damage to machine parts.
Key words：ai deep learning; Robot collision; Pre estimation controller; Human computer interaction; PC sensing device; PyTorch framework; Activation function;
收稿日期：2022-12-15
基金项目：
