IST Lunch Bunch

Tuesday November 1, 2011 12:00 PM

Motion control of autonomous systems: reconfigurable spacecraft and fast air vehicles

Speaker: Marin Kobilarov, Control & Dynamical Systems, Caltech
Location: Annenberg 105
The talk examines challenges in the motion control of autonomous systems operating in constrained environments. We discuss two applications under development: fast unmanned ground and aerial robotic vehicles navigating optimally through an obstacle terrain; autonomous reconfiguration of distributed spacecraft subject to orbital environment constraints. In particular, the Autonomous Assembly of Reconfigurable Space Telescope (AAReST) project, recently initiated by Caltech/JPL and the University of Surrey, will be introduced. AAReST aims to demonstrate the construction of a large telescope from multiple distributed "mirrorcraft" (a small cubesat-based spacecraft equipped with propulsion and carrying a segment of the telescope mirror) capable of autonomous orbital navigation and docking. Planned for launch by 2015, this proof-of-concept mission proposes a paradigm shift in space science observation in terms of unprecedented scalability and reduced cost. Improved computational theory and control algorithms required for such systems, and more generally for nonlinear control systems found in robotics and aerospace, will be discussed. Simulating and optimizing their motion is addressed in terms of geometric computational optimal control methods on manifolds. An algorithmic optimization framework for systems with symmetries such as rigid multi-body assemblies and robotic locomotion systems is developed. In addition, the complexity of high-dimensional systems operating among environmental obstacles, is addressed in terms of randomized sampling-based methods for global motion planning. These approximation methods are based on adaptive sampling of progressively optimal trajectories by optimally exploiting the collected information about the problem. This new approach to the control of complex systems results in not only the ability to quickly find a feasible solution but to find an approximately optimal one with high probability. We will show promising ongoing application of these algorithms to autonomous vehicles.
Series IST Lunch Bunch

Contact: Sydney Garstang at x2813 sydney@caltech.edu
For more information visit: http://www.cs.caltech.edu/seminars/lunch_bunch.html