Matrix sensitivity based real-time iterative feedback control method for leg hydraulic drive system of legged robot

• A novel matrix sensitivity based real-time iterative feedback control (MRIFC) method is proposed. • The MRIFC gradually optimizes the feedforward compensation term through online real-time iterative feedback control of predicted tracking errors, effectively mitigating the im-pact of noise and disturbance signals on HDU control performance. • A sensitivity control approach is innovatively developed. Unlike sensitivity analysis methods, sensitivity control accurately maps predicted tracking errors to current control inputs for feedforward compensation. By computing feedforward compensation gain coefficients in real-time via sensitivity factors, it achieves variable-gain feedforward compensation control. • The real-time iterative feedback control method proposed in this paper is implemented within mathematical models, ensuring compliance with real-time control requirements of practical systems. The high-precision control of the hydraulic drive unit (HDU) is the key to realize the highly agile motion of the hydraulic legged robot. However, the system uncertainty and coupling characteristics seriously restrict the control performance of HDU. In this paper, a novel matrix sensitivity-based real-time iterative feedback control (MRIFC) method is proposed. Firstly, a complete sensitivity control theory is proposed, and the mathematical representation between prediction error and input is established by matrix sensitivity. Secondly, the online iterative feedback control theory is proposed. The feedforward compensation gain coefficient is constructed according to the sensitivity factor. The feedforward compensation term is gradually optimized by online iterative feedback control in the mathematical model. This mechanism fundamentally resolves the impact of operating conditions, external disturbance forces, and system non-linearity on feedforward compensation gain. It prevents system control oscillations caused by high gain and exhibits strong adaptability to varying operating conditions. Finally, the effectiveness of MRIFC is verified by a large number of experiments. The experimental results show that the control accuracy of MRIFC is improved by 27.3 % on average compared with the contrast control strategy. The MRIFC has high control accuracy and robustness, which lays a technical foundation for the high dynamic stability control of hydraulic legged robot.