A Separation Force Position Hybrid Control for Quadruped Robots Based on Optimization Strategy

Objective: Aiming at insufficient stability of quadruped robot in diagonal trot gait and low real-time efficiency of mainstream controllers, this paper proposes a decoupled Separate optimal force mixing based on quadratic programming optimization. It intends to suppress body attitude fluctuation, smooth the effect of Optimal force distribution at the foot end, and eliminate switching instability of traditional hybrid control. Methods: For supporting legs, virtual feedback forces/torques are constructed at the centroid to realize Optimal force distribution at the foot end, and foot force distribution is converted into a constrained QP problem to obtain optimal ground reaction forces. Hip and knee joints adopt torque control, while lateral swing joints use independent position control to realize Separate optimal force mixing without state selector. Swing legs combine cubic polynomial trajectory planning and Virtual Model Control for trajectory tracking. Poincaré mapping theory is adopted to verify system stability, and comparative simulations are carried out against VMC, FPHC and MPC in MATLAB/Simulink. Results: The robot’s roll angle fluctuates within ±1°, pitch angle varies ±0.5° with negligible yaw drift, outperforming contrast algorithms. The scheme of Optimal force distribution at the foot end cuts foot force peaks and mutations to reduce joint impact. Analyzed by Poincaré mapping, all Floquet multipliers fall within the unit circle to confirm periodic orbital stability. The algorithm’s solving time stays below 3 ms, greatly faster than MPC with a maximum latency of 25 ms. Conclusion: The proposed Separate optimal force mixing realizes full decoupling of force and position control on joints of the quadruped robot to avoid switching instability. The Optimal force distribution at the foot end based on QP improves walking stability, and Poincaré mapping theory verifies gait stability. Compared with mainstream control algorithms, the combination with Virtual Model Control delivers better attitude tracking and faster computation, meeting high real-time demands for diagonal trot locomotion of quadruped robots.