A resource allocation method and system based on multi-agent

Abstract

This application discloses a resource allocation method and system based on multi-agents, including: S1 The controller selects a sub-strategy; S2 Obtains the observation information of the environment for the selected sub-strategy, and obtains action rewards for executing actions; S3 Repeats S2, and the processing unit Using the obtained multiple action rewards, use the reinforcement learning algorithm to update the parameters of the selected sub-strategy; the S4 processing unit updates the utility average value of the selected sub-strategy with a distributed algorithm; the S5 processing unit uses the utility average value and the policy reward feedback formula, Determine the policy reward of the selected sub-strategy, save it to the storage unit; S6 controller selects a new sub-strategy, and executes S2 to S5 in a loop until the number of cycles reaches the threshold number of times; S7 processing unit according to each strategy corresponding to each sub-strategy Reward, update the controller parameters with a reinforcement learning algorithm. Hierarchical reinforcement learning models enable agents to quickly adapt to complex environments. Using multiple sub-strategies and cooperating with other agents to determine policy rewards has strong versatility.

Description

A method and system for resource allocation based on multi-agent technical field [0001] The application relates to the field of artificial intelligence, in particular to a multi-agent-based resource allocation method and system. Background technique Reinforcement learning (Reinforcement Learning) is a branch of machine learning, which mainly includes four Elements: Agent, Environment State, Action, and Reward. Rewards are provided by the environment A quantifiable scalar feedback signal to the agent, used to evaluate the agent's performance at a certain time step (Time Step). Doing the action is good or bad. The goal of reinforcement learning is to get the most cumulative reward. Reinforcement learning is the control of a An agent acting autonomously in the environment continuously improves its behavior through interaction with the environment. Reinforcement learning problems include Learn what to do and how to map the environment into action for maximum reward...

AI Technical Summary

Problems solved by technology

This method uses linear programming or game theory calculations, but it cannot adapt to complex environments and has a small scope of application.

In addition, some multi-agent reinforcement learning methods have also been applied to resource allocation, such as computing resources, network resources, and logistics systems, etc. However, these methods rely on domain expertise and cannot be generalized to general situations

Images