A Massage Manipulator Training Method Based on Deep Reinforcement Learning

Abstract

The invention discloses a massage manipulator training method based on deep reinforcement learning, which solves the problems in the prior art that the action of the massage manipulator is only in a simulated state and the massage action is not accurate enough, and can enhance the skill of the massage manipulator and provide professional and accurate training. Massage and reduce the fatigue effect of manual massage; its technical scheme is: collecting action and pressure data, processing the data, constructing a reference action set and a reference pressure set, and setting the comfort range of pressure values; combining the data, reference The action and reference pressure are input into the neural network for prediction and decision-making, and the neural network is executed to output the action value and pressure value corresponding to the decision, and compared with the reference action and pressure value comfort range; after the set conditions are met, the trained network and massage manipulator connected to the control system.

Description

technical field [0001] The invention relates to the field of manipulators, in particular to a massage manipulator training method based on deep reinforcement learning. Background technique [0002] Nowadays, the mechanical equipment used for massage is not diversified. Most of them are single-function or multi-functional massagers, massage chairs, etc., with few movements and mechanization. It is difficult to grasp the use of force, and cannot provide users with more comfortable, More professional service. Manual massage movements are delicate and soft, especially professional massage, which is highly skilled and skillful. However, the number of professional masseuses is small, and they cannot provide services anytime and anywhere, and the cost is relatively high, so they cannot meet the needs of ordinary people. [0003] With the development of artificial intelligence and the continuous improvement of productivity requirements, industrial robotic arms have been used in mo...

AI Technical Summary

Problems solved by technology

Due to the high-dimensional sample complexity and other physical limitations of the reinforcement learning algorithm, the combination of deep learning and reinforcement learning greatly reduces the dimension and complexity of the data, but it is currently only in the simulation state and cannot be fully applied to the actual situation.

[0004] At present, the main points to overcome in the field of manipulator control are path planning and trajectory planning. However, the imitation of manipulator actions, especially the use of deep reinforcement learning methods for manipulator action imitation is very rare, and it is very difficult to implement.

Images