Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Terrain-Attentive Learning for Efficient 6-DoF Kinodynamic Modeling on Vertically Challenging Terrain

Aniket Datar, Chenhui Pan, Mohammad Nazeri, Anuj Pokhrel, Xuesu Xiao

TL;DR

This paper proposes a 6-DoF kinodynamics learning approach that is attentive only to the specific underlying terrain critical to the current vehicle-terrain interaction, so that it can be efficiently queried in real-time motion planners onboard small robots.

Abstract

Wheeled robots have recently demonstrated superior mechanical capability to traverse vertically challenging terrain (e.g., extremely rugged boulders comparable in size to the vehicles themselves). Negotiating such terrain introduces significant variations of vehicle pose in all six Degrees-of-Freedom (DoFs), leading to imbalanced contact forces, varying momentum, and chassis deformation due to non-rigid tires and suspensions. To autonomously navigate on vertically challenging terrain, all these factors need to be efficiently reasoned within limited onboard computation and strict real-time constraints. In this paper, we propose a 6-DoF kinodynamics learning approach that is attentive only to the specific underlying terrain critical to the current vehicle-terrain interaction, so that it can be efficiently queried in real-time motion planners onboard small robots. Physical experiment results show our Terrain-Attentive Learning demonstrates on average 51.1% reduction in model prediction error among all 6 DoFs compared to a state-of-the-art model for vertically challenging terrain.

Terrain-Attentive Learning for Efficient 6-DoF Kinodynamic Modeling on Vertically Challenging Terrain

TL;DR

This paper proposes a 6-DoF kinodynamics learning approach that is attentive only to the specific underlying terrain critical to the current vehicle-terrain interaction, so that it can be efficiently queried in real-time motion planners onboard small robots.

Abstract

Wheeled robots have recently demonstrated superior mechanical capability to traverse vertically challenging terrain (e.g., extremely rugged boulders comparable in size to the vehicles themselves). Negotiating such terrain introduces significant variations of vehicle pose in all six Degrees-of-Freedom (DoFs), leading to imbalanced contact forces, varying momentum, and chassis deformation due to non-rigid tires and suspensions. To autonomously navigate on vertically challenging terrain, all these factors need to be efficiently reasoned within limited onboard computation and strict real-time constraints. In this paper, we propose a 6-DoF kinodynamics learning approach that is attentive only to the specific underlying terrain critical to the current vehicle-terrain interaction, so that it can be efficiently queried in real-time motion planners onboard small robots. Physical experiment results show our Terrain-Attentive Learning demonstrates on average 51.1% reduction in model prediction error among all 6 DoFs compared to a state-of-the-art model for vertically challenging terrain.
Paper Structure (19 sections, 3 equations, 5 figures, 1 table)

This paper contains 19 sections, 3 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Wheeled Robot Kinodynamic Modeling
  4. Off-Road Navigation
  5. Learning-Based Mobility
  6. Approach
  7. Problem Formulation
  8. Data-Driven Kinodynamics
  9. Terrain-Attentive Learning (tal)
  10. Implementations
  11. Terrain Attentive Learning
  12. 6-DoF Kinodynamics Learning
  13. Experiments
  14. Robot, Testbed, and Data
  15. Trajectory Prediction Visualization
  16. ...and 4 more sections

Figures (5)

  • Figure 1: Two Sets of 6-DoF Kinodynamic Trajectory Predictions by tal and wmvctdatar2023learning Compared to Ground Truth.
  • Figure 2: Terrain-Attentive Learning (tal, Left) and 6-DoF Kinodynamics Learning (Right) Architecture: Flame and temperature denote training and frozen parameters respectively.
  • Figure 3: 6-DoF Vehicle Trajectories of tal, wmvct, and Ground Truth with Increasing Horizon: tal closely matches Ground Truth even with a long horizon, while wmvct significantly diverges.
  • Figure 4: Model Prediction Error of tal and wmvct: Average One-Step 6-DoF Positional and Angular Error (Left); Prediction Error vs. Prediction Step (Middle and Right). tal achieves lower prediction error and variance than wmvct in all cases.
  • Figure 5: Diverse 6-DoF Vehicle States in the Dataset.