Shape-Space Graph: Fast and Collision-Free Path Planning for Soft Robots

This repository contains the code accompanying the paper:
Shape-Space Graph: Fast and Collision-Free Path Planning for Soft Robots

by Carina Veil, Moritz Flaschel, Ellen Kuhl.

Contact: [email protected]

If you use this code in your research, please cite us: https://arxiv.org/abs/2510.03547

---

Overview

This project presents a graph-based path planning tool for a soft robotic arm. Using a biomechanical model inspired by morphoelasticity and active filament theory, it precomputes a shape library and constructs a k-nearest neighbor graph in shape space, where each node corresponds to a mechanically accurate and physically valid robot shape. Signed distance functions prune nodes and edges colliding with obstacles, while multi-objective edge costs combining geometric distance and actuation effort enable energy-efficient planning with collision avoidance. Shortest paths are efficiently found with Dijkstra’s algorithm using the precomputed graph.

Features

• Load toy data of soft robotic arm shapes and activations
• Define obstacles with flexible SDFs including boxes and cylinders
• Compute clearance and collision checks for graph edges
• Build weighted k-nearest neighbor graphs in shape space
• Plan routes through shape waypoints using Dijkstra’s algorithm
• Visualize shape sequences, activations, and obstacles in 3D plots

---

File Structure

• main.py: Entry script to run the path planning
• edge_utils.py: Functions to compute shape distances, edge weights, and collision checks
• graph_utils.py: Graph construction, nearest neighbors, shortest path, and waypoint planning
• sdf_utils.py: Signed distance functions for obstacles and scenes
• plotting_utils.py: 3D plotting utilities for shapes, activations, and obstacles

---