Adaptive parallel reflex- and decoupled CPG-based control for complex bipedal locomotion

The achievement of adaptive, stable, and robust locomotion and dealing with asymmetrical conditions for bipedal robots remain a challenging problem. To address the problem, this paper introduces adaptive parallel reflex- and decoupled central pattern generator (CPG)-based control for a planar bipedal robot. The control has modular structure consisting of two parallel modules that work together. Firstly, as the main controller, the reflex-based control module inspired by an agonist–antagonist model, utilizes proprioceptive sensory feedback to adaptively generate various stable gaits. In parallel, as an auxiliary controller, the decoupled CPG-based control units individually governing the robot legs have the ability to learn the generated gaits in an online manner. Using the proposed framework, our study shows that this real-time control approach contributes to stable gait generation with robustness toward sensory feedback malfunction and adaptability to deal with environmental and morphological changes. Herein this study, we demonstrate the planar bipedal robot control functionality on a variable speed treadmill, dealing with asymmetric conditions such as weight imbalance and asymmetrical elastic resistance in the legs. However, the approach does not require robot kinematic and dynamic models as well as an environmental model and is therefore flexible. As such, it can be used as a basis for controlling other bipedal locomotion systems, like lower-limb exoskeletons.