Autonomous equipment decision control method based on distributed reinforcement learning

Abstract

The invention discloses an autonomous equipment decision control method based on distributed reinforcement learning. The method comprises a training environment building step, a distributed training step and a decision model outputting step. A sampling node, a cache node and a training node are configured in a distributed cluster, an autonomous equipment simulation environment is packaged into a container mirror image, and virtual resources are allocated. A distributed agent training system is started, training end model parameters are initialized, the parameters are sent to a cache node and then forwarded to a sampling node, after the interaction process of an agent and a simulation environment is completed, data are returned to the training node, and the model parameters are updated by using a deorbit deep reinforcement learning algorithm. And after the distributed training is completed, exporting the model from the system, and switching to a rapid reasoning mode for intelligent decision-making. According to the method, training can be carried out on a large-scale distributed cluster in the implementation process, hardware resources can be fully utilized, and meanwhile the bandwidth requirement can be remarkably reduced for improvement of a communication mode.

Description

technical field [0001] The invention relates to a decision-making control method for autonomous equipment based on distributed reinforcement learning, and belongs to the technical field of autonomous equipment control and distributed systems. Background technique [0002] In reality, the realization of autonomous equipment decision control based on reinforcement learning has the problem of large data demand. With the trend of integration and complexity of research on decision-making control of autonomous equipment, it is difficult to achieve results in a limited time by only using a single-machine method for intelligent agent training of decision-making control of autonomous equipment. [0003] In the existing reinforcement learning solutions, the training of autonomous equipment decision control agents is severely restricted by limited computing resources. When faced with more complex and more real problem scenarios, the ability of deep reinforcement learning models that ca...

AI Technical Summary

Problems solved by technology

With the trend of comprehensive and complex research on decision-making control of autonomous equipment, it is difficult to achieve results in a limited time for autonomous equipment decision-making control agent training using only a stand-alone method.

[0003] In existing reinforcement learning solutions, the training of autonomous device decision-making control agents is severely constrained by limited computing resources. When faced with more complex and more realistic problem scenarios, the capabilities of deep reinforcement learning models that can be trained are limited. Often only solve single-field problems such as obstacle avoidance, path planning, and dynamic control

Images