Self-adaptive dynamic programming optimal control method and system based on co-state assistance

Abstract

The invention discloses a self-adaptive dynamic programming optimal control method based on co-state assistance. The method comprises the following steps: step 1, constructing a state variable, a control variable, a transfer environment, a utility function and a cost function of a system; step 2, constructing an execution network, a co-state network and an evaluation network of self-adaptive dynamic programming based on co-state assistance; step 3, carrying out a strategy evaluation process of self-adaptive dynamic planning based on co-state assistance; step 4, carrying out a strategy improvement process of self-adaptive dynamic planning based on co-state assistance; and step 5, repeating the steps 3-4 until the loop iteration frequency i is greater than the specified frequency or the cost function difference value abs(Ji+1(xt, ut, [Lambda]t)-Ji(xt, ut, [Lambda]t)) after two adjacent external loops is less than a set value. According to the self-adaptive dynamic programming optimal control method based on co-state assistance, the derivative of the cost function to the state quantity serves as an auxiliary variable when the cost function is estimated, meanwhile, the process of dynamic modeling on a complex system is avoided, and the solving precision of the complex dynamic system is improved.

Description

technical field [0001] The invention belongs to the technical field of intelligent control, and in particular relates to an adaptive dynamic programming optimal control method and system based on co-state assistance. Background technique [0002] Complex dynamic systems are ubiquitous in many research fields, such as vehicle systems, underwater robot systems, chemical process systems, energy management systems, etc. With the development of science and technology, people have put forward higher and higher requirements on the performance of these control systems. requirements. Optimal control is to seek a control quantity on the basis of ensuring that the control system can achieve the goal stably, so that the given system cost function reaches the maximum value (or minimum value). However, in practical applications, complex systems usually have characteristics such as high nonlinearity, unknown dynamic characteristics, and model uncertainty, which will make it difficult to e...

AI Technical Summary

Problems solved by technology

Among them, the patent "An Approximate Dynamic Programming Optimization Control Method for Underwater Thermal Glider Attitude Adjustment" owned by Shanghai Maritime University discloses a two-input and two-output decoupling control that is not based on models. Update the control law online, but the cost function cannot guarantee accurate evaluation during the online training process, so it may get worse control effect during the training process

Images