Robot reinforcement learning method based on adaptive model

Abstract

The invention provides a robot reinforcement learning method based on an adaptive model. According to the robot reinforcement learning method, the step of model self-adaption is added while an environment model is learned normally, and therefore the effect of improving the accuracy of virtual data can be achieved. Specifically, when the model is of a neural network structure, the model adaptivelyimproves the accuracy of the model on virtual data by reducing the feature distribution of real data and virtual data on a network hidden layer. Experiments prove that in the fields of robot control and the like, compared with a previous model-based reinforcement learning method, the method has higher sampling efficiency and final performance.

Description

technical field [0001] The invention relates to the field of artificial intelligence, in particular to a robot reinforcement learning method. Background technique [0002] In the field of robot control, we define the state as the position and speed of each part of the robot, and the action as the force applied to each part. The goal is to make the robot move as far as possible without falling over, and at the same time make the robot move as far as possible. The control force should be as small as possible, and the reinforcement learning method can be used to complete the above tasks. [0003] Reinforcement learning can be divided into model-free reinforcement learning and model-based reinforcement learning according to whether the environment is modeled. Among them, model-free reinforcement learning uses (state, action, next action, reward) tuple data sampled in the real environment to directly train a policy or value function, while model-based reinforcement learning meth...

AI Technical Summary

Problems solved by technology

In order to solve the problem of data distribution shift in domain adaptation, an effective method is to learn features with invariant properties

Images