考虑多关节时滞的液压挖掘机轨迹跟踪研究

Trajectory tracking of hydraulic excavators considering the multijoint time delay

为了减小多关节系统的时滞特性对液压挖掘机工作装置轨迹跟踪精度的影响,本文设计了考虑时滞的长短期记忆-广义回归级联网络和一种基于该网络模型的多关节控制系统。利用几何法分析闭式链运动状态,建立以摇臂关节为末端反馈关节的含闭式链的工作装置运动学逆解,通过4-3-3-3-4多项式轨迹规划算法在关节空间上生成轨迹。结合LSTM的时序处理能力和GRNN的强非线性映射能力,以比例阀信号和关节转角的各阶延迟以及目标转角作为网络输入特征,通过多关节联动辨识建立起输出信号和输入特征的映射关系模型,将关节模型作为逆控制器实现多关节轨迹跟踪。经AMESim-Simulink仿真验证:在空载或变载荷作用下,与不考虑时滞的LSTM系统比,考虑时滞的多关节控制系统能更准确、更迅速、更平滑地跟踪规划轨迹,误差小,并且系统具有较强鲁棒性。

To reduce the influence of the time-delay characteristics of the multijoint system on the trajectory tracking accuracy of the working device on a hydraulic excavator,an architecture for long-short-term memory-generalized re-gression neural networks considering time delays(TD-GR-LSTM)and a multijoint control system based on this net-work model are presented.First,the motion states of the closed chain by the geometric method are analyzed,and the inverse kinematics solution of the working device is established with a closed chain that uses the rocker arm joint as the end feedback joint.The trajectory of joint space is generated by the 4-3-3-3-4 polynomial trajectory plan-ning algorithm.Next,combining with the timing processing ability of LSTM and the strong nonlinear mapping abili-ty of GRNN and taking the target rotation angle and each delay of the proportional valve signal and joint rotation an-gle as the network input characteristics,the mapping relationship between output signals and input characteristics is established through multijoint coordinated identification.Then,the multijoint model is used as the inverse controller to realize multijoint trajectory tracking.Finally,AMESim-Simulink cosimulation on a medium excavator shows that under the no-load or variable load of different frequencies,compared with the LSTM control system not considering time delay,the developed multijoint control system considering time delays can track the planned trajectory faster and more accurately and smoothly,with smaller tracking error and strong robustness.