Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Neural Configuration Distance Function for Continuum Robot Control

Kehan Long, Hardik Parwana, Georgios Fainekos, Bardh Hoxha, Hideki Okamoto, Nikolay Atanasov

TL;DR

This paper addresses the challenge of real-time safe planning for deformable continuum robots by learning a Neural Configuration Euclidean Distance Function (N-CEDF) that represents each link's shape as a distance field and fuses them through the robot's kinematic chain. The method leverages per-link distance functions, trained with a safety-oriented loss, to enable efficient collision checking from point-cloud data and integrates them into a Model Predictive Path Integral (MPPI) controller for safe, dynamic navigation. Key contributions include the N-CEDF formulation, a loss function that discourages distance overestimation, and extensive simulations demonstrating accurate shape modeling, scalable planning across multiple-link robots, and favorable real-time performance in cluttered and dynamic environments. The approach offers practical impact for deploying continuum robots in complex settings where traditional point-cloud or rigid-body abstractions struggle, enabling safer, faster motion planning with limited sensing.

Abstract

This paper presents a novel method for modeling the shape of a continuum robot as a Neural Configuration Euclidean Distance Function (N-CEDF). By learning separate distance fields for each link and combining them through the kinematics chain, the learned N-CEDF provides an accurate and computationally efficient representation of the robot's shape. The key advantage of a distance function representation of a continuum robot is that it enables efficient collision checking for motion planning in dynamic and cluttered environments, even with point-cloud observations. We integrate the N-CEDF into a Model Predictive Path Integral (MPPI) controller to generate safe trajectories for multi-segment continuum robots. The proposed approach is validated for continuum robots with various links in several simulated environments with static and dynamic obstacles.

Neural Configuration Distance Function for Continuum Robot Control

TL;DR

This paper addresses the challenge of real-time safe planning for deformable continuum robots by learning a Neural Configuration Euclidean Distance Function (N-CEDF) that represents each link's shape as a distance field and fuses them through the robot's kinematic chain. The method leverages per-link distance functions, trained with a safety-oriented loss, to enable efficient collision checking from point-cloud data and integrates them into a Model Predictive Path Integral (MPPI) controller for safe, dynamic navigation. Key contributions include the N-CEDF formulation, a loss function that discourages distance overestimation, and extensive simulations demonstrating accurate shape modeling, scalable planning across multiple-link robots, and favorable real-time performance in cluttered and dynamic environments. The approach offers practical impact for deploying continuum robots in complex settings where traditional point-cloud or rigid-body abstractions struggle, enabling safer, faster motion planning with limited sensing.

Abstract

This paper presents a novel method for modeling the shape of a continuum robot as a Neural Configuration Euclidean Distance Function (N-CEDF). By learning separate distance fields for each link and combining them through the kinematics chain, the learned N-CEDF provides an accurate and computationally efficient representation of the robot's shape. The key advantage of a distance function representation of a continuum robot is that it enables efficient collision checking for motion planning in dynamic and cluttered environments, even with point-cloud observations. We integrate the N-CEDF into a Model Predictive Path Integral (MPPI) controller to generate safe trajectories for multi-segment continuum robots. The proposed approach is validated for continuum robots with various links in several simulated environments with static and dynamic obstacles.
Paper Structure (14 sections, 15 equations, 6 figures, 4 tables)

This paper contains 14 sections, 15 equations, 6 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. Neural Configuration Euclidean Distance Function
  4. Configuration EDF for Continuum Robot
  5. Data Preparation and Loss Function
  6. Safe Motion Planning and Control
  7. MPPI for Continuum Robot Control
  8. Cost Function Design
  9. Evaluation
  10. Simulation Setup
  11. Neural Network Architecture Comparison
  12. Navigating Dynamic Environments
  13. Navigating with Point-cloud Data
  14. Conclusion

Figures (6)

  • Figure 1: (a) A single continuum robot link with parameters: arc lengths $l_{1}, l_{2}, l_{3}$, backbone length $L$, bending angle $\theta$, and bending plane angle $\varphi$. (b) A 4-link continuum robot with a specific configuration in a dynamic environment with spherical obstacles and an end-effector goal (blue star).
  • Figure 2: Visualization of the N-CEDF for a continuum robot link.
  • Figure 3: End-effector distance to goal and distances from robot to obstacles.
  • Figure 4: Safe navigation of a 4-link continuum robot. The blue star denotes the goal, the colored shapes denote the static obstacles, and the red spheres denote the dynamic obstacles. The MPPI planned trajectory of its end-effector is shown in blue dots.
  • Figure 5: 5-link continuum robot navigation in a cluttered environment.
  • ...and 1 more figures