Dynamic Mobility via Cellular Decompositions of Coordination Spaces
Final rept. 1 Oct 2011 - 30 Jun 2012
CARNEGIE-MELLON UNIV PITTSBURGH PA
Pagination or Media Count:
To accomplish feats of rapid, nimble locomotion, the ground mobile robots of tomorrow must exhibit great dexterity and dynamic mobility. As their complexity increases, however, the need for principled approaches to coordinating a robots actuators becomes apparent. We have begun to pursue a new formalism for solving such problems, making use of Algebraic Topology and classical group theory to replace traditional methods of combinatorial search and optimization with a computationally easier and conceptually more straightforward approach that identifies and exploits symmetries in a system. In our approach, we decompose a system into cells that index symmetric configurations. For a multilegged robot negotiating its way through a rubble-strewn terrain, the robot must choose between a number of alternative ways to move its legs. We reduce this problem to changes on a much simpler space of gait timing, using a cell complex to identify all possible gaits, discretizing the space while also providing a roadmap for gait transitions. By studying these representations---and the cellular decompositions that arise---we develop novel approaches to the fundamental control problems necessary for achieving robotic mobility.
- Physical Chemistry
- Surface Transportation and Equipment