Reproduce “RL-Based Coverage Path Planning for Deformable Objects on 3D Surfaces” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yuhang Zhang, Jinming Ma, Feng Wu
An RL-based framework is introduced for complete coverage path planning using deformable objects on 3D surfaces. This approach employs harmonic UV mapping to simplify 3D states to 2D for efficient learning, achieving shorter paths and competitive coverage compared to baseline methods in simulation, with successful real-world deployment on a robotic arm.
Key insights
- Dimensionality reduction through UV mapping allows an RL agent to efficiently learn complex 3D surface coverage paths using 2D map representations.
- Integrating low-cost force feedback from a physics simulator (MuJoCo) effectively provides the necessary interaction data for training, circumventing real-world sensory challenges.
- A reward function that prioritizes newly covered areas and penalizes an increase in coverage map's Total Variation guides the agent to generate both efficient and thorough coverage paths.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.