TALK: Adaptive Model-Predictive Motion Planning for Autonomous Robots in Complex Environments
Speaker: Thomas Howard, NASA JPL
Date: Wednesday, June 27 2012
Time: 10:00AM to 11:00AM
Location: Patil/Kiva Seminar Room (32-G449)
Host: Daniela Rus, MIT
Contact: Mieke Moran, 617-253-5817, firstname.lastname@example.orgRelevant URL:
Autonomous robots must reason about their surroundings to operate intelligently in the natural world. Provided infinite time and flawless perception, robots could exploit detailed models of their environment interaction and completely explore the space of possible decisions to determine the optimal course of action. Practical applications of robotics however have restricted resources, limited decision time, and imperfect model information. In this talk I will present my research in model-predictive motion planning algorithms for autonomous robots that operate in challenging, cluttered, and/or complex environments. These methods are centered on two core ideas. First, a real-time trajectory generation that is agnostic to the underlying mechanics of robot environment interaction can be used to generate motion planning search spaces for autonomous robots that are feasible, expressive, and efficient to search. Second, provided a computationally efficient technique for repairing connectivity in recombinant motion planning graphs, the local mapping between discretized and continuous representations can be relaxed to improve the relative optimality of generated motions. I will demonstrate generality of these approaches by discussing applications in planetary rovers, field robots, autonomous automobiles, mobile manipulators and robotic torsos and describe future directions for this work.
Thomas Howard is a Research Technologist in the Robotics Software Systems Group at the Jet Propulsion Laboratory. He received his Ph.D. in Robotics from Carnegie Mellon University in 2009 and earned B.S. degrees in Mechanical Engineering and Electrical and Computer Engineering from the University of Rochester in 2004. His work centers on the development of robust motion planning, navigation, and control algorithms that are applicable to a wide spectrum of autonomous systems. At Carnegie Mellon University he focused on the development of model-predictive trajectory generation and mobile robot navigation algorithms for planetary rovers, field robots, autonomous automobiles, and mobile manipulators. He was a member of the Tartan Racing motion planning team and applied his local motion planning trajectory generation algorithms on Boss, winner of the 2007 DARPA Urban Challenge. While at the Jet Propulsion Laboratory he has led research tasks on perception and pose estimation in poorly illuminated environments and mobile robot search space design. He is currently the motion planning lead for the Jet Propulsion Laboratory and California Institute of Technology team on the software track of the DARPA Autonomous Robotic Manipulation program and a member of the Mars Science Laboratory flight software team working on autonomous navigation. Thomas has authored or co-authored four journal articles, nineteen conference papers, and a book chapter in robot motion planning, navigation, control, and simulation and twice been appointed as a Lecturer in Mechanical Engineering at the California Institute of Technology where he helps teach the class Advanced Robotics: Navigation and Vision.
See other events happening in June 2012