粒子群算法优化BP神经网络的变载荷自平衡控制系统

上海工程技术大学 机械工程学院,上海 201620

时变负载; 自平衡; 粒子群算法; PID神经网络

Variable load self-balancing control system basedon PSO-optimized BP neural network
NI Shou-bin,CHENG Wu-shan

(School of Mechanical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)

time-varying load; self-balancing; particle swarm optimization; PID neural network

DOI: 10.13800/j.cnki.xakjdxxb.2017.0624

备注

为解决常规PID控制难以在具有时变负载的自平衡系统中实时、精确调节负载的变化,在BP神经网络基础上,利用粒子群算法(PSO)优化BP神经网络,将神经网络的收敛速度进一步提高,并将算法应用到二轮平衡车控制系统中,对二轮平衡车进行动力学建模,介绍系统的结构、原理与实验方法,搭建二轮平衡车实验平台进行了施加突变负载情况下的试验验证。利用二轮平衡车实验平台车身上的姿态传感器得到车体倾斜输出角度,对比施加突变负载前后以及神经网络优化前后的车体倾斜输出角度变化。结果 表明:粒子群算法(PSO)优化BP神经网络技术能够满足变负载二轮自平衡车控制的要求,实现了自平衡车的动态自平衡,提高了抗干扰能力,验证了优化算法在自平衡、抗外部干扰和缩短调整时间上的优势。

In order to solve the difficulty in conventional PID control with time-varying load balance system in real-time and precise adjustment of the load changes,on the base of BP neural network and using particle swarm algorithm(PSO)optimized BP neural network,the convergence speed of neural network was improved,and the algorithm was applied to the two-wheel balance vehicle control system.A dynamic model of two-wheel balance vehicle was established.We introduce the structure,principle and experimental method of the system,build the experimental platform for the two-wheel balance vehicle on the test case of mutation load,get the output angle of tilt body posture sensor using two-wheel balance vehicle experimental platform on the vehicle body,contrast output tilt angle changes before and after applied load and neural network optimization.The test results shows that the particle swarm algorithm(PSO)-optimized BP neural network technology meets the varying load in the two wheel self-balancing vehicle control,the dynamic self-balancing of the self-balancing vehicle is realized,and the anti-interference capability is improved.The advantages of the optimized algorithm in self-balancing,anti external interference and shortening the adjustment time are verified.