基于粒子群算法的混联机构神经网络自适应反演控制

针对含有不匹配干扰的混联机构轨迹跟踪控制问题，提出了一种极限学习机与自适应反演控制相结合的控制策略。在对干扰进行分析的基础上，分别采用两个极限学习机网络对系统中的匹配和不匹配干扰进行逼近和补偿。基于Lyapunov函数稳定性设计了混联机构的控制律与自适应律，实现混联机构的轨迹跟踪控制。由于控制器可调参数较多，采用粒子群算法进行控制器参数的寻优整定。仿真结果表明，所提出的控制方法具有良好的轨迹跟踪精度和鲁棒性。

Neural Network Adaptive Backstepping Control of Hybrid Mechanism Based on PSO

Aiming at the trajectory tracking control problem of the hybrid mechanism with mismatched disturbance, a control strategy combining extreme learning machine and adaptive backstepping control was proposed. Considering the hybrid mechanism containing the characteristics of drive motor, adaptive control with backstepping method was used to design the control strategy in stages. Based on the disturbance analysis, the conveying mechanism modeling error, friction, load and external random disturbance, and motor voltage disturbance were taken as matched disturbance and mismatched disturbance were two lumped disturbance terms. Since the mismatched disturbance cannot be eliminated directly by the feedback controller, two ELM networks were used to perform on-line approximation respectively, and perform feedforward compensation in the designed backstepping controller. According to the stability theory of Lyapunov function, the control rate and adaptive rate of the hybrid mechanism were designed. The simulation results showed that the method effectively eliminated the influence of mismatch disturbance in the system and realized the trajectory tracking control of the hybrid mechanism. In addition, because the neural network adaptive inversion controller of the hybrid mechanism contained many adjustable parameters such as inversion stabilization coefficients and adaptive parameters, the particle swarm algorithm was used to optimize and set the controller parameters. The system error, output error, controller output and rise time were used as the objective function construction conditions, and the optimal parameters of the controller were obtained through 150 iterations of optimization. Neural network adaptive backstepping controller without PSO-optimize and the PD controller were simulated as a comparison. The simulation results showed that the neural network adaptive backstepping controller of the hybrid mechanism based on PSO optimization had excellent tracking accuracy and system robustness.