Pump-Pressure-Compensation-Based Adaptive Neural Torque Control of a Hydraulic Excavator with Open Center Valves

To autonomously fulfil some typical working tasks, such as the side wall excavation and the pressing work, the swing torque of a hydraulic excavator should be controlled. However, it is difficult to develop a useful torque controller for a hydraulic excavator due to: 1) the system has a special structure, which results in that the system control gains are zero under certain conditions; 2) the existing hydraulic force/torque control methods are mainly developed based on a nominal system model, which is hard to be obtained for a hydraulic excavator. In this paper, a novel pump-pressure-compensation-based adaptive neural torque controller is proposed for a hydraulic excavator. To guarantee that the control gains are not zero and the system is controllable, a novel hydraulic structure is proposed and the pump pressure is compensated for. In addition, to improve the torque tracking control performance, an adaptive neural torque controller is synthesized by utilizing the universal approximation ability of the neural network. An echo state network is adopted to approximate a lumped uncertain system function and it has a guaranteed approximation with online-updated weights. In addition, to deal with the lumped uncertainties resulting from the neural network approximation error, a nonlinear robust term is introduced. The influence of the lumped uncertainties is handled by using a novel parameter adaption law, which estimates and updates the upper bound of the lumped uncertainties online. With the proposed method, the torque tracking error convergence can be theoretically proved. The superior performance of the proposed method is validated through comparative experiments.