Reducing the Complexity of Robot Motion Planning: Algorithms and Applications in Bipedal Locomotion

Bipedal locomotion, while captivating in its resemblance to human movement, presents a significant challenge within the field of robotics. The inherent multi-dimensionality and dynamic complexity of bipedal robots have posed substantial obstacles to traditional control methods, often limiting their real-world applications. This talk will delve into the complexity of bipedal motion planning, presenting algorithmic approaches that aim to simplify control and unlock the practical potential of bipedal robots. We examine a range of model-based and learning-based approaches to develop reduced-dimensional representations for a better understanding of robot behaviors and their interactions with the environment. A significant focus of our research is on designing stable walking gaits that account for the robots’ natural dynamics and environmental uncertainties. By developing computationally efficient optimization and learning frameworks, we experimentally demonstrated the feasibility of dynamic and robust walking with challenging humanoid robots and realized essential capabilities such as accurate velocity tracking, traversing complex terrain, navigating dynamic obstacles, and exhibiting stability under external perturbations. Furthermore, we explore the potential for knowledge transfer from humanoid robots to lower-limb exoskeletons, potentially enabling restoration of mobility for individuals with paraplegia and enhancing workplace productivity and safety.

Speaker: Ayonga Hereid, Ohio State University