Clothoidal Local Path Template for Intention Estimation by Assistive Mobile Robots

  • Kevin Denis KU Leuven
  • Johan Philips KU Leuven
  • Herman Bruyninckx KU Leuven
  • Eric Demeester KU Leuven

Abstract

This work proposes and evaluates various improvements to a circular local path template (LPT) that we have designed in the past to estimate driver intents and to provide navigation assistance to wheelchair drivers. This LPT may also be useful in other mobile robotics applications such as for intuitive teleoperated control, for fast collision checking in path planning or for obstacle avoidance algorithms. The LPT consists of a large but fixed set of paths in the mobile robot’s local neighborhood. Based on an efficient look-up table, the LPT paths’ lengths are adjusted such that they are collisionfree. However, experiments have shown that in dense environments insufficient circular paths are found, which impedes correct intention estimation and thus navigation assistance. In this work, the use of clothoidal paths rather than circular paths is evaluated. This substantially improves the capability to find complex paths in dense areas. Furthermore, we adapt the LPT to deal with dynamic obstacles of random shape using motion prediction estimates of these objects.

Published
2019-07-06