Learning Control Sets for Lattice Planners скачать в хорошем качестве

Learning Control Sets for Lattice Planners

Title: Learning Control Sets for Lattice Planners from User Preferences Authors: Alexander Botros, Nils Wilde, Stephen L. Smith University of Waterloo, Canada Abstract: We investigate the design of a lattice-based motion planner that can capture user preferences. Given demonstrated trajectories we estimate a user cost function, and compute an optimal motion primitives that reflect the learned user preference. Further, we restrict the number of motion primitives in the lattice to ensure strong performance during motion planning. We show that this problem is optimally solved by applying a separation principle: First, we find the best estimate of the user cost function given data, then an optimal control set is computed given that estimate. We evaluate our work in simulations for an autonomous robot in a 4-dimensional spatio-temporal lattice and show that the proposed approach is suitable to replicate the demonstrated behaviour while enjoying substantially increased performance. Keywords: Robot Motion Planning, Human-Robot Interaction