Research On Posture Control Of Two-wheeled Mobile Car Based On Adaptive And Improved Particle Swarm Optimization Algorithm

Two-wheeled mobile car is a typical nonlinear and under-actuated system, and it is an ideal platform for verify kinds of control algorithm. The thesis focus on the problem of mobile car’s posture control. The thesis analysis the advantages and disadvantages of hierarchical sliding-mode controller at first, then design the adaptive fuzzy sliding-mode controller and adaptive neural network sliding controller; Meanwhile, the thesis using adaptive particle swarm algorithm to optimize the parameters of controller, to overcome the arbitrariness of the controller parameter initialization. Effectiveness and superiority of the algorithm is verified by theoretical analysis and experimental results. The thesis’s main work and innovation points as follows:(1)Because the system exist parameter perturbation and external disturbance, which seriously affect the performance of traditional sliding mode controller, so introduce adaptive fuzzy method to sliding mode control algorithm, and use fuzzy logic system to predict system functions and external disturbance. At the same time, the thesis design adaptive law by Lyapunov stability to adjust system parameters on line.. The stability of the system verified by theoretical analysis, the effectiveness and superiority of the control algorithm verified by the results of simulation experiment.(2)Because the neural network can approximate any nonlinear function with any precision, the thesis take advantage of this characteristic to design adaptive neural-network sliding controller, and use this controller to countervail the total amount of system uncertainty. Meanwhile, predicting controller’s gain on line by neural network to eliminate chattering of controller’s output. The stability of the system verified by theoretical analysis, the effectiveness and superiority of the control algorithm verified by the results of simulation experiment.(3)In order to overcome the arbitrariness of the neural network parameter initialization and particle swarm algorithm’s deficiencies, the thesis use adaptive particle swarm algorithm to optimize the parameter of controller, this way can improved the performance of controller.