Random nonlinear multi-agent reinforcement learning optimization formation control method

Abstract

The invention relates to the technical field of adaptive non-linear control, particularly discloses a reinforcement learning optimization formation control method for random non-linear multi-agent, and designs an adaptive identifier for estimating unknown random power based on the function approximation capability of a neural network. Then, optimization control is obtained by constructing an evaluation network and executing reinforcement learning through the network; in the invention, the reinforcement learning algorithm is obtained by executing a gradient descent method on a simple positive function, and the function is designed according to the partial derivative of an HJB equation, so that the optimal control is simpler than the traditional method, and can be more conveniently applied to a random nonlinear multi-agent system; and finally, verifying from two aspects of theorem proving and computer simulation, so that the proposed optimization method can realize an expected target.

Description

technical field [0001] The invention relates to the technical field of adaptive nonlinear control, in particular to a reinforcement learning optimization formation control method of random nonlinear multi-agents. Background technique [0002] The control design of stochastic systems is a very challenging problem because its differentiation involves not only stochastic disturbances, but also Hessian terms in stability analysis. With the development of control theory, some popular nonlinear control techniques, such as Sontag's stabilization formula, backstepping technique and adaptive observer, have been extended to stochastic systems. It is worth mentioning that some optimal control methods have also been extended to stochastic nonlinear systems. However, since stochastic multi-agent control requires state coupling, which brings difficulties to the optimal design, the optimal formation control of stochastic multi-agent systems is rarely reported. The main reason is that tra...

AI Technical Summary

Problems solved by technology

However, the optimal formation control of stochastic multi-agent systems is seldom reported due to the state coupling required for stochastic multi-agent control, which brings difficulties to the optimal design.

The main reason is that traditional reinforcement learning optimization methods are either complex algorithms or require some strict conditions, such as: continuous incentives, they are difficult to apply and extend to stochastic multi-agent system control

Images