Active Admittance Control With Iterative Learning for General-Purpose Contact-Rich Manipulation

摘要

Robotic force interactions are unavoidable in various operational contexts. However, ensuring that robots can effectively handle diverse tasks involving force control remains a formidable challenge. This article addresses the need for reproducible interaction tasks and the lack of a comprehensive force control framework for multitask scenarios. To tackle these issues, we propose a new combined framework which introduces iterative learning control (ILC) strategy into the conventional admittance control. The framework uses a new admittance parameter tuning approach, which endows the algorithm with automatic parameters tuning ability. We evaluate the proposed framework using four representative manipulation tasks to assess its consistency and generalizability. Experimental results confirm the effectiveness of the framework, demonstrating an average improvement of 98.21% and 91.52% in root mean square error (RMSE) compared to conventional admittance control and model-free adaptive control, respectively.