Abstract:
Bipedal robots have long held the promise of walking around in the human world the dynamic way not approachable by traditional wheel robots. The rapid development of mechanical and actuation capabilities of modern robots has already made more dynamic behaviors possible. However, exploiting the full potential of these machines to demonstrate efficient and agile motions as humans remains a difficult task. One of the reasons why robots are not able to do what human can do is the physical complexity challenges our best computational tool for planning and controlling the type of dynamic behaviors of these robots. Now it is the situation where us robotic scientists have the opportunity to raise the capability of these systems to change the way they move in the world and allow them to do many of things we know that they are able to do in principle. 

In this talk, the speaker will present a dynamics based motion planning framework for bipedal robots by effectively unifying the formal control theory with modern optimization methods. By decoupling the complex full-order dynamics constraints and exploiting the sparsity structure of the nonlinear optimization problem, these methods are capable of generating energy-efficient dynamic gait trajectories for legged robots in a computationally efficient way. A drawback of trajectories is that they provide little information on how to respond to a disturbance. To address this shortcoming, Supervised Machine Learning is used to extract a low-dimensional state-variable realization of the open-loop trajectories. We then use the special structure of mechanical models associated with bipedal robots to embed the low-dimensional model in the original model in such a manner that the desired walking motions are locally exponentially stable. The effectiveness of the proposed framework is demonstrated by the experimental realization of dynamic and energy efficient walking with a full-size humanoid robot at the DARPA Robotics Challenges in 2015, and fully autonomous walking with an actuated lower-limb exoskeleton. Finally, the speaker will present a brief overview of his future research vision.

Bio:
Ayonga Hereid (阿永嘎) is a postdoctoral research fellow at the University of Michigan, Ann Arbor, where he is a member of Dynamic Legged Locomotion Lab working with Professor Jessy Grizzle. He received his Ph.D. degree in Mechanical Engineering from the Georgia Institute of Technology in 2016, and his bachelor’s and master’s degrees in Mechanical Engineering from the Zhejiang University in 2007 and 2010, respectively. His current research interests focus on control and motion planning for bipedal robots and exoskeletons, with a particular emphasis on developing optimization-based feedback control solutions for robotic and hybrid dynamical systems. His work in dynamic bipedal locomotion received the DENSO Best Student Paper award at HSCC (ACM International Conference on Hybrid Systems: Computation and Control) in 2014 and was nominated to the finalist for the Best Paper award at ICRA (IEEE International Conference on Robotics and Automation) in 2016.