Optimization method of multi-agent system consistency problem based on sub-graph processing

Abstract

The invention discloses a multi-agent system consistency problem optimization method based on sub-graph processing, and the method is characterized in that the method comprises the following steps: 1)constructing a graph signal model of a multi-agent system; the method comprises the following steps of (1) solving a problem to be solved, (2) adding an auxiliary constraint condition to the problemto be solved, (3) carrying out local inversion in sub-graphs, (4) carrying out fusion averaging between the sub-graphs, and (5) iteratively eliminating errors, and the method is capable of reducing operation complexity, achieving consistent averaging, high in convergence speed and low in requirements for computing capacity and storage capacity of a single node of a system.

Description

technical field [0001] The invention relates to the technical field of multi-agent systems, in particular to an optimization method for the consistency problem of a multi-agent system based on subgraph processing. Background technique [0002] Multi-agent systems are widely used in flight formations, robot formations, traffic control systems, sensor networks, and many other fields. A single agent in a multi-agent system has limited sensing, communication, and computing capabilities, while the entire multi-agent system uses the cooperation between agents to complete high-complexity tasks. As a carrier of distributed information processing, multi-agent systems have been widely studied in the fields of computer science, communication engineering, biological intelligence, and automatic control. In the cooperative control problem of multi-agent systems, achieving the required consistency is the prerequisite for agents to coordinate and cooperate, so the consistency problem has r...

AI Technical Summary

Problems solved by technology

Xiao and Boyd first proposed a deterministic Gossip algorithm. In this algorithm, each node exchanges information with its neighbor nodes at the same time, and then calculates the weighted average value between itself and all neighbors. When the weight matrix composed of weighted coefficients is a consistent average matrix, the information of all nodes will converge to the average consistency; Nedic and Ozdaglar refer to the idea of ​​deterministic Gossip algorithm, and propose a distributed gradient descent method. The gradient descent method minimizes the local cost function to speed up the consistent convergence speed. This kind of gradient-based distributed optimization algorithm is simple to implement, but the convergence speed is slow

To this end, some scholars have proposed a distributed algorithm based on dual theory, such as Terelius et al. proposed a distributed alternating direction method of multipliers (Alternating Direction Method of Multipliers, ADMM), the main idea of ​​ADMM is to introduce dual variables, the original optimization problem It is decomposed into two sub-problems. Each node needs to update its own state and calculate the update of the dual variable. Compared with the general gradient algorithm, the distributed ADMM algorithm improves the convergence speed, but requires higher storage and computing capabilities of the nodes.

Images