Understanding Environment-Adaptive Force Control of Series Elastic Actuators

Series elastic actuators (SEAs) have become fundamental components in robots that physically interact with the surrounding word and with humans. Force control of SEAs is indeed an active area of research. This work refines and improves the stability analysis of the environment-adaptive force controller we previously proposed and proves asymptotic convergence in several cases of practical interest. In particular, we theoretically motivate certain effects which are not adequately explained or covered in our previous work. The analysis reveals an interesting generalization property of the approach, achieved by using a simple and generic model to account for very different environments including stiff contacts, purely inertial loads and soft materials. This is allowed by model parameters with inter-changeable physical meaning: the same parameter describes different physical properties or variables, depending on the kind of interacting environment. In this light, the proposed adaptive controller adapts not only to parameters values, but also to parameter meaning. The analysis also shows that as the environment stiffness decreases, the convergence precision may degrade and a sliding-mode robustification is proposed to overcome the issue. Simulations and experiments are conducted to validate the control convergence in different environments, showing agreement with the theoretical expectation.