Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SIL-RRT*: Learning Sampling Distribution through Self Imitation Learning

Xuzhe Dang, Stefan Edelkamp

TL;DR

SIL-RRT* is a novel learning-based motion planning algorithm that extends the RRT* algorithm by using a deep neural network to predict a distribution for sampling at each iteration, establishing that it can efficiently solve high-dimensional motion planning problems with fewer samples than traditional sampling-based algorithms.

Abstract

Efficiently finding safe and feasible trajectories for mobile objects is a critical field in robotics and computer science. In this paper, we propose SIL-RRT*, a novel learning-based motion planning algorithm that extends the RRT* algorithm by using a deep neural network to predict a distribution for sampling at each iteration. We evaluate SIL-RRT* on various 2D and 3D environments and establish that it can efficiently solve high-dimensional motion planning problems with fewer samples than traditional sampling-based algorithms. Moreover, SIL-RRT* is able to scale to more complex environments, making it a promising approach for solving challenging robotic motion planning problems.

SIL-RRT*: Learning Sampling Distribution through Self Imitation Learning

TL;DR

SIL-RRT* is a novel learning-based motion planning algorithm that extends the RRT* algorithm by using a deep neural network to predict a distribution for sampling at each iteration, establishing that it can efficiently solve high-dimensional motion planning problems with fewer samples than traditional sampling-based algorithms.

Abstract

Efficiently finding safe and feasible trajectories for mobile objects is a critical field in robotics and computer science. In this paper, we propose SIL-RRT*, a novel learning-based motion planning algorithm that extends the RRT* algorithm by using a deep neural network to predict a distribution for sampling at each iteration. We evaluate SIL-RRT* on various 2D and 3D environments and establish that it can efficiently solve high-dimensional motion planning problems with fewer samples than traditional sampling-based algorithms. Moreover, SIL-RRT* is able to scale to more complex environments, making it a promising approach for solving challenging robotic motion planning problems.

Paper Structure

This paper contains 14 sections, 6 equations, 4 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Problem Statement
  5. Transformer
  6. Imitation Learning
  7. Method
  8. Sampler Model Architecture
  9. Weighted Self Imitation Learning
  10. Experiment Results
  11. Data Collection
  12. Results
  13. Conclusion and Limitation
  14. Acknowledgements

Figures (4)

  • Figure 1: Illustration of a 2D state space scenario and two contrasting methods for representing the state space using point clouds. In our experiment, we opted to sample point clouds from the surface of obstacles rather than their interior regions.
  • Figure 2: Architecture of Sampler Model
  • Figure 3: Architecture of Estimator
  • Figure 4: Examples of paths found by RRT*, SIL-RRT* w/o and with WSIL algorithms in various scenarios. Left column shows the scenario and the following ones the trajectories found, which are best visualized magnified and in colors on screen.