Mechanical arm dense object temperature priority grabbing method based on deep reinforcement learning

Abstract

The invention discloses a mechanical arm dense object temperature priority grabbing method based on deep reinforcement learning. The method comprises the following steps of S1, constructing a mechanical arm working space, and constructing a real-time state; S2, preprocessing state information; S3, carrying out forward transmission on the preprocessed information through a Q network to obtain a corresponding Q value; S4, instructing a mechanical arm to act according to the Q value and an epsilon-greedy strategy, and obtaining rewards through a reward function; S5, continuously updating the weight through a target Q network to realize training of the Q network; and S6, recording related data in the training process and a model finally formed after training to obtain an optimal grabbing strategy of the mechanical arm. The method has the advantages that the method is specifically applied to the grabbing scenes where objects have irregular shapes and are densely stacked and temperature factors need to be preferentially considered; the mechanical arm action is designed according to the deep reinforcement learning algorithm, so that the grabbing performance of the mechanical arm is improved; and an infrared image is introduced, so that the mechanical arm has the capacity of preferentially grabbing objects with higher temperature.

Description

technical field [0001] The present invention relates to a temperature-priority grabbing method for dense objects with a manipulator based on deep reinforcement learning. Deep reinforcement learning is applied to the grasping task of a manipulator, and pushing and grasping are put into a joint action within a reinforcement learning framework to promote Promote grasping and set temperature rewards, so that the robotic arm can better grasp dense objects and have the ability to preferentially grasp high-temperature objects. Background technique [0002] At present, the application and function of robotic arms are becoming more and more perfect; with the rapid development of robotic arm technology, robotic arms have been widely used in industrial tasks such as handling, palletizing, cutting, welding, etc., which not only liberate manpower, but also improve industrial production. The efficiency and quality of the robot arm; among them, the grasping task of the robotic arm is the b...

AI Technical Summary

Problems solved by technology

[0004] The characteristics of the above grasping scenarios have brought difficulties to the grasping work of the robotic arm; the actual modeling process of the model-based method is usually complicated, and it is difficult to apply to unstructured scenarios; Grasping has some effect, but it is not effective for densely stacked objects, and it cannot give priority to hazard conditions

Images