基於軟體驅動實現足型機器人全伺服運動控制器

本文提出一種以軟體驅動的全伺服運動控制系統架構，應用於足型機器人的精密運動控制。系統採用分層式設計，其中上層控制器負責步態規劃與運動決策，下層控制器則負責執行各關節的伺服驅動。本文在足型機器人控制方面比較傳統控制策略與現代機器學習策略的差異，並設計一種混合運動控制迴路，來結合前饋扭力補償與回饋控制以提升系統穩定性與響應。整體架構基於即時作業系統與EtherCAT通訊協議來實現多關節同步控制。實驗結果顯示此控制架構能夠在1 kHz循環下穩定足型機器人的步態表現，達成高精度與高即時性的控制效能。

This article presents a software-defined full-servo motion control architecture tailored for high-precision locomotion of lagged robots. The proposed system adopts a hierarchical control scheme wherein the high-level controller performs gait generation and motion planning, while the low-level controller manages real-time actuation of joint-level servo motors. A comparative analysis is conducted between conventional control algorithms and modern machine learning-based techniques, leading to the development of a hybrid control loop that combines feedforward torque estimation with closed-loop feedback compensation. The implementation is based on real-time operation system with the EtherCAT, enabling deterministic and synchronized multi-joint control at a loop frequency of 1 kHz. Experimental evaluations demonstrate that the proposed architecture effectively maintains gait stability and delivers high-fidelity, real-time motion control performance, validating its suitability for dynamic lagged locomotion.