气动肌肉关节的神经元PID轨迹跟踪控制

Neuron PID Trajectory Tracking Control of Pneumatic Muscle Joints

摘要:

气动肌肉关节具有柔顺性,使角度跟踪控制困难;通过建立气动肌肉关节静态平衡方程,分析了气压与摆角间非线性关系。为克服单神经元PID控制算法收敛速度慢、精度低的缺点,基于Hebb学习规则,从反馈周期、比例增益系数和衰减因子三方面改进算法；搭建了实物平台,通过实验和对比分析可得:(1)适当减小闭环反馈周期,可提高跟踪精度,但闭环反馈周期太小会引起震荡；(2)将比例增益系数定义为误差的连续有界Sigmoid函数,可增强算法的自适应能力,提高跟踪精度；(3)增加衰减因子,模拟人的遗忘学习机制,可提高权值的收敛速度。

Abstract:

Pneumatic muscle joints have the character of flexibility, meanwhile it increases angle tracking control difficulty. Static equilibrium equation of pneumatic muscle joint was established to analyze the nonlinear relationship between pressure and angle. To overcome single neuron adaptive PID control algorithm’s weaknesses of slow convergence rate and low control accuracy, three improved methods were proposed from three aspects of feedback cyclicality,proportional gain coefficient and attenuation factor. All of these were based on Hebb learning rule. From the experiments we have done on physical platform and contrastive analysis of the results ,we can concluded that (1)Decreasing closed-loop feedback cyclicality can improve tracking accuracy .But it will cause concussion if closed-loop feedback cyclicality is too small.(2) A sigmoid function associated with error is proposed to define proportional gain coefficient. This method enhances the adaptive ability and improves the tracking accuracy. (3) An attenuation factor can be added to simulate people’s forgotten learning mechanism, which can improve convergence rate of weight numbers.