Graph Neural Network Based Method for Path Planning Problem

TL;DR

This paper introduces a Graph Neural Network-based approach to path planning that reduces collision detection and enhances efficiency in high-dimensional robotic environments.

Contribution

A novel GNN-based neural network model that uses environment maps to guide path planning and minimize collision checks.

Findings

Significantly reduces collision detection in simulations

Improves path planning speed in high-dimensional spaces

Effective in both simulated and real-world tests

Abstract

Sampling-based path planning is a widely used method in robotics, particularly in high-dimensional state space. Among the whole process of the path planning, collision detection is the most time-consuming operation. In this paper, we propose a learning-based path planning method that aims to reduce the number of collision detection. We develop an efficient neural network model based on Graph Neural Networks (GNN) and use the environment map as input. The model outputs weights for each neighbor based on the input and current vertex information, which are used to guide the planner in avoiding obstacles. We evaluate the proposed method's efficiency through simulated random worlds and real-world experiments, respectively. The results demonstrate that the proposed method significantly reduces the number of collision detection and improves the path planning speed in high-dimensional environments.