Sampling-based motion planning for uncertain high-dimensional systems via adaptive control

C. K. Verginis, D. V. Dimarogonas, and L. E. Kavraki, “Sampling-based motion planning for uncertain high-dimensional systems via adaptive control,” 2020 Workshop on Algorithmic Foundations of Robotics, 2020.

Abstract

This paper considers the problem of safe motion planning for high-dimensional holonomic robots with 2nd-order uncertain dynam- ics. We integrate sampling-based motion planning techniques with tra- ditional adaptive feedback control and address difficulties encountered in planning for such systems. More specifically, we develop a feedback control scheme that tracks a given reference trajectory within certain bounds, while simultaneously compensating for potential uncertainty in the dynamical parameters of the robot (masses, moments of inertia) and external disturbances. Employing this result, we are able to cast the problem of kinodynamic motion planning to a geometric one, which can be usually solved more efficiently since it does not take into account the robot dynamics (a.k.a. differential constraints). Intuitively, we use a geometric planner to obtain a high-level safe path for the robot, and a low-level adaptive feedback control algorithm to execute it while tak- ing into account the robot dynamics. Experimental results validate the proposed approach.

PDF preprint: http://kavrakilab.org/publications/verginis2020.pdf