Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Planning with Learned Subgoals Selected by Temporal Information

Xi Huang, Gergely Sóti, Christoph Ledermann, Björn Hein, Torsten Kröger

TL;DR

This paper proposes a method that leverages a generative model to decompose a complex planning problem into small manageable ones by incrementally generating subgoals given the current planning context by taking into account the temporal information.

Abstract

Path planning in a changing environment is a challenging task in robotics, as moving objects impose time-dependent constraints. Recent planning methods primarily focus on the spatial aspects, lacking the capability to directly incorporate time constraints. In this paper, we propose a method that leverages a generative model to decompose a complex planning problem into small manageable ones by incrementally generating subgoals given the current planning context. Then, we take into account the temporal information and use learned time estimators based on different statistic distributions to examine and select the generated subgoal candidates. Experiments show that planning from the current robot state to the selected subgoal can satisfy the given time-dependent constraints while being goal-oriented.

Planning with Learned Subgoals Selected by Temporal Information

TL;DR

This paper proposes a method that leverages a generative model to decompose a complex planning problem into small manageable ones by incrementally generating subgoals given the current planning context by taking into account the temporal information.

Abstract

Path planning in a changing environment is a challenging task in robotics, as moving objects impose time-dependent constraints. Recent planning methods primarily focus on the spatial aspects, lacking the capability to directly incorporate time constraints. In this paper, we propose a method that leverages a generative model to decompose a complex planning problem into small manageable ones by incrementally generating subgoals given the current planning context. Then, we take into account the temporal information and use learned time estimators based on different statistic distributions to examine and select the generated subgoal candidates. Experiments show that planning from the current robot state to the selected subgoal can satisfy the given time-dependent constraints while being goal-oriented.

Paper Structure

This paper contains 17 sections, 9 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Learning heuristics
  4. Learning distributions
  5. Neural planners
  6. Method
  7. Dataset
  8. Generating Spatial Subgoals
  9. Learning Temporal Distributions
  10. Subgoal Selection based on Temporal Estimation
  11. Start-to-sample metric
  12. Goal-to-sample metric
  13. Planning range shaping
  14. Experiments and Results
  15. Quantitative Evaluation
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Planning with learned subgoals: with the estimated probabilistic distribution of planning time from the start to subgoal candidates (orange), and from the goal to subgoal candidates (turquoise), candidate 2 is selected. Planning ranges are adapted to the blue dashed bounding box after the selection for planning efficiency.
  • Figure 2: The PLS architecture. The green arrows show the direction of gradient descent. During training, the backpropagation of the time estimator does not reach the encoding blocks.
  • Figure 3: An example of the dataset. The start and goal configurations are marked in cyan and green, respectively. The subgoals, namely ground truths are shown in yellow.
  • Figure 4: Samples of planning time described by normal and log-normal distributions. Red: statically determined. Green: estimated by a model.
  • Figure 5: Two robot arms experiment. Gray: current G-robot statek. Green: final goal. Yellow: selected subgoals by PLS. Pink: P-robot as a moving object running a fixed trajectory.