Optimizing Gaits for Coverage on Lie Groups

B Bittner, Shai Revzen

Dynamic Walking, 2019

Abstract

Individual performance criteria, such as robustness and efficiency, often drive behavior selection in mobile robotic platforms. However, navigational capabilities can rely heavily on the coordination of all behaviors available to the robot, rather than the performance of one behavior. Here, we studied how a collection of availabile gaits (cyclic internal motions) can be related to the variability and density of external motions that a system can achieve. We attempted to capture this capability, termed ”coverage”, with an associative fast-to-compute cost function that can be used to simultaneously optimize a collection of gaits. The cost function can be weighted such that users may specify the importance of various classes of external motions. A resulting optimization framework produced three gaits for a 3-link Purcell swimmer, locally maximizing coverage on the space of planar rigid body motions.

BibTeX

@conference{bittner-2019-dw,
  author = {Bittner, B and Revzen, Shai},
  title = {Optimizing Gaits for Coverage on Lie Groups},
  booktitle = {Dynamic Walking},
  year = {2019},
  month = jun,
  organization = {Dynamic Walking},
  url = {https://docs.google.com/document/d/155IGklc5VWjCga3UaZyJ66UeFiV2_94IlAl5AmnqK7Y/view},
}