Autonomous Grasping Method of Robotic Arm Based on Deep Reinforcement Learning and Dynamic Motion Primitives

Abstract

The invention discloses an autonomous grasping method of a mechanical arm based on deep reinforcement learning and dynamic motion primitives, which includes the following steps: Step 1: Install a camera image component to ensure that the recognition area is not blocked, and preprocess the image of the captured target area , and sent to the deep reinforcement learning agent as state information; Step 2: Construct a local policy proximal optimization training model based on the state and deep reinforcement learning principles; Step 3: Construct a new dynamic motion primitive and imitation learning Hybrid motion primitive model; Step 4: Based on the model, train the robotic arm to autonomously grasp objects. The present invention can effectively solve the problem of unsmooth joint motion of the mechanical arm based on traditional deep reinforcement learning. By combining the dynamic motion primitive algorithm, the meta-parameter learning problem is transformed into a reinforcement learning problem, and the training method of deep reinforcement learning can be used to make the mechanical arm Complete autonomous grabbing tasks.

Description

technical field [0001] The invention relates to the technical field of a robotic arm and a deep reinforcement learning training system, in particular to an autonomous grasping method of a robotic arm based on deep reinforcement learning and dynamic motion primitives. Background technique [0002] At present, the research of robotics technology has shifted from traditional mechanical dynamics to intelligent control, especially after comprehensively absorbing research results in the fields of control theory, artificial neural network and machine learning, robotics has gradually become an artificial intelligence One of the cores of the field. As one of the research hotspots in the field of machine learning in recent years, deep reinforcement learning has achieved rich results in both theoretical research and practical applications. However, a good deep reinforcement learning algorithm is not enough for robots to solve real-life problems. This is because the control strategy o...

AI Technical Summary

Problems solved by technology

Since the joint motion of the robot is driven and controlled by the motor, if the motion trajectory (angle trajectory, angular velocity trajectory and angular acceleration trajectory) of the motor has large fluctuations, the driving torque of the motor will also have large fluctuations at this time. even large mutation values, which can easily cause damage to robot joints

Images