Online adjustment method for control parameters of autonomous aircraft based on MCMC optimized Q learning

Abstract

The invention discloses an online adjustment method for the control parameters of an autonomous aircraft based on MCMC optimized Q learning, which includes the following steps: first, doing statisticsof possible changes of the PID control parameters of an aircraft according to the actual situation to get an action set of parameter adjustment, and initializing the PID control parameters accordingto the aircraft control experience; then, randomly selecting an action to act on the autonomous aircraft, carrying out sampling through an MCMC algorithm according to the function value Q* of each action obtained in a Q learning algorithm to get an action to be taken next moment, and adjusting the learning factor l in the Q learning algorithm through an SPSA step size adjustment algorithm with thepassage of time; and finally, getting the optimal control parameters in the current environment through repeated adjustment of the control parameters. The problems of overshoot and delay in the navigation process of an autonomous aircraft are solved, and an autonomous aircraft can quickly adapt to the change of the environment and quickly and steadily reach the destination.

Description

technical field [0001] The invention belongs to the field of on-line adjustment of control parameters of an autonomous aircraft, and specifically relates to a method for adjusting the control parameters of an autonomous aircraft. Background technique [0002] The autonomous navigation of the aircraft means that the aircraft reaches the destination designated by humans on the water surface, and then independently plans the path of travel, and finally reaches the destination through continuous self-regulation. It has important application value in water quality inspection and water surface patrol. [0003] At present, the traditional autonomous aircraft adopts the fixed PID parameter method, which uses fixed aircraft control parameters, and the parameters are obtained from a large number of aircraft autonomous navigation engineering project experience. When the fixed control parameters are not suitable for the current environment, it will cause problems of overshoot and respo...

AI Technical Summary

Problems solved by technology

When the fixed control parameters are not suitable for the current environment, it will cause problems of overshoot and response delay for the autonomous navigation of the aircraft, especially in the case of changing environments, the fixed control parameters may have a better response to individual environmental states , but it cannot meet all environmental conditions. When the environment changes, it is necessary to manually change the control parameters of the aircraft, which is not convenient for the use of the aircraft.

[0004] There are also some methods that use fuzzy algorithm and annealing algorithm to adjust the control parameters of aircraft. These methods introduce the automatic adjustment mechanism of control parameters to a certain extent. Still can't solve the problem of quickly adjusting the control parameters of the autonomous vehicle to the optimal value

Images