A visibility-based approach to computing non-deterministic bouncing strategies. (September 2021)

• Record Type: Journal Article
• Authors: Nilles, Alexandra Q; Ren, Yingying; Becerra, Israel; LaValle, Steven M
• Other Names: Morales Marco guest-editor.; Tapia Lydia guest-editor.; Sánchez-Ante Gildardo guest-editor.; Hutchinson Seth guest-editor.
• Is Part Of: International journal of robotics research. Volume 40:Number 10/11(2021)
• Page Start: 1196
• Page End: 1211
• Publication Date: 2021-09
• Subjects: Underactuated robots -- dynamics -- motion control -- motion planning; Robots -- Periodicals; Robots, Industrial -- Periodicals; 629.89205
• DOI: 10.1177/0278364921992788
• Languages: English
• ISSNs: 0278-3649

Abstract

Inspired by motion patterns of some commercially available mobile robots, we investigate the power of robots that move forward in straight lines until colliding with an environment boundary, at which point they can rotate in place and move forward again; we visualize this as the robot "bouncing" off boundaries. We define bounce rules governing how the robot should reorient after reaching a boundary, such as reorienting relative to its heading prior to collision, or relative to the normal of the boundary. We then generate plans as sequences of rules, using the bounce visibility graph generated from a polygonal environment definition, while assuming we have unavoidable non-determinism in our actuation. Our planner can be queried to determine the feasibility of tasks such as reaching goal sets and patrolling (repeatedly visiting a sequence of goals). If the task is found feasible, the planner provides a sequence of non-deterministic interaction rules, which also provide information on how precisely the robot must execute the plan to succeed. We also show how to compute stable cyclic trajectories and use these to limit uncertainty in the robot's position.