15:30 – 15:55 Patrick Wensing — MIT, USA.

Title: “Real-time planning and control for the MIT Cheetah”

Abstract: The challenge of dynamic locomotion is well captured by the fact that high-speed gaits are never instantaneously balanced in a traditional sense. This challenge motivates the need for real-time predictive capabilities to unlock new levels of dynamic mobility in our robots. Towards this aim, the talk will focus on model-predictive control (MPC) for planning and stabilization of dynamic gaits in the MIT Cheetah 2. An MPC approach will be described for bounding in the sagittal plane, which has enabled autonomous jumping over obstacles up to 80% of leg length in hardware. To extend this framework for more general 3D gaits, a new approach of policy-regularized MPC (PR-MPC) will be discussed. This method has empirically been found to improve both the solution speed and outcomes of solving non-convex MPC optimization problems. PR-MPC is envisioned for application to a new robot, the MIT Cheetah 3, currently under final construction. The talk will conclude with a short description of this platform and ongoing efforts to validate the PR-MPC approach.

Bio: Patrick Wensing is a Postdoctoral Associate in the Biomimetic Robotics Laboratory within the Department of Mechanical Engineering at MIT. He received his Ph.D. in Electrical and Computer Engineering from The Ohio State University in 2014. He was awarded an NSF Graduate Research Fellowship for his dissertation research on balance control strategies for humanoid robots. At MIT, the results of his postdoctoral work on the MIT Cheetah robot have received considerable publicity worldwide.