Events Calendar

CS Colloq: Matt Zucker

Title: Combining Planning and Optimization for Rough Terrain LocomotionAbstract:Motion planning for rough-terrain legged robots is a difficult task, not only due to the high dimensionality of robot configuration spaces, but also due to the variety of kinematic, dynamic, and collision constraints which need to be met at all times. While producing optimal walking behavior is desirable, searching the space of all posible robot motions remains intractible for non-trival robotic systems. In this talk, I describe a hierarchy of planning and optimization algorithms that decomposes the planning problem into a sequence of decisions which can be efficiently solved in order to produce real-time locomotion over rough terrain. My software architecture has been successfully used over the past year to guide the LittleDog quadruped robot over a variety of terrain types. Beyond this specific software architecture, I will also discuss the ways in which machine learning and optimization techniques can increase the speed and quality of motion planning algorithms, and highlight lessons learned on how to decompose a high-level planning task into a tractable set of sub-problems.Bio:Matt Zucker is a Ph.D. candidate at the Robotics Institute at Carnegie Mellon University, where he works on motion planning for high degree-of-freedom robotic platforms. His research focuses on leveraging numerical optimization and machine learning techniques in order to improve planning speed and quality. Before graduate school, Matt worked from 2000-2005 writing software for autonomous underwater vehicles at Bluefin Robotics Corporation in Cambridge, MA. He expects to graduate from the Robotics Institute in the summer of 2010.Host: Prof. Stefan Schaal

Location: Seaver Science Library (SSL) - 150

Audiences: Everyone Is Invited

Contact: CS Front Desk