Research On Automatic Leveling Systemvia Neural Network Decoupling Control

Platform automatic leveling technology is widely applied in military,industry,scientific research and other fields.It acts a significant role in supporting the national economic construction and the social development.Key performance indexes for evaluating the automatic leveling control system are leveling speed,leveling accuracy,stability,etc.Because of the platform is a nonlinear multi parameters coupling system,it is difficult for traditional method to obtain excellent control effect.Therefor this paper studies on modeling,simulation and intelligent control of platform system in order to improve its leveling speed,leveling accuracy and stability.Firstly,this paper choices four legs structure platform as the research object,and analyzes the features among position error control leveling method,angle error control leveling method and Inverse system decoupling leveling method.Then,a compound leveling method was proposed in this paper,which combined with the highest point fixed method and the inverse system decoupling method.Secondly,this paper builds the simulation models of platform system and leg system in Matlab/Simulink based on the analysis of mathematical models of platform system and its subsystem.On this basis,this paper uses classical PID controllers to control the platform,the results of simulation indicate that the leveling speed is slow due to the incomplete decoupling control of the classical PID controllers.In order to solve the problem,a kind of PID neural network(PIDNN)that modified on its structure and learning algorithm was proposed in this paper.Simulation results show that the leveling speed of modified PIDNN is obviously faster than classical PID controller.However the PIDNN is easy to fall into the partial optimum during its online learning process that caused a large steady-state error.Therefor the PIDNN controller is difficult to feed the demand of high leveling precision.Finally,in order to eliminate the large steady-state error of PIDNN decoupling controller.This paper presents a kind of modified cuckoo search algorithm to optimize the initial weights of PIDNN,which combined with particle swarm optimization(PSO)algorithm,genetic algorithm(GA)and adaptive operator execution.According to the contrast experiments of standard test function,the results show that the optimization search performance of MCS is better than CS and PSO.And the simulation results of whole system indicated that the optimized PIDNN was not only maintained the rapidleveling speed,but also eliminated the steady-state error,and improved the leveling performance of platform.