Predicting Sample Collision with Neural Networks

TL;DR

This paper introduces a neural network-based framework combining a Contractive AutoEncoder and a Multilayer Perceptron to rapidly predict robot collision states, significantly reducing computation in motion planning for high-dimensional spaces.

Contribution

It presents a novel approach integrating CAE and MLP for fast collision prediction, improving efficiency and generalization in sampling-based motion planning.

Findings

Framework is computationally efficient across various problems.

Framework generalizes well to new workspaces.

Significantly reduces collision detection time.

Abstract

Many state-of-art robotics applications require fast and efficient motion planning algorithms. Existing motion planning methods become less effective as the dimensionality of the robot and its workspace increases, especially the computational cost of collision detection routines. In this work, we present a framework to address the cost of expensive primitive operations in sampling-based motion planning. This framework determines the validity of a sample robot configuration through a novel combination of a Contractive AutoEncoder (CAE), which captures a occupancy grids representation of the robot's workspace, and a Multilayer Perceptron, which efficiently predicts the collision state of the robot from the CAE and the robot's configuration. We evaluate our framework on multiple planning problems with a variety of robots in 2D and 3D workspaces. The results show that (1) the framework is computationally efficient in all investigated problems, and (2) the framework generalizes well to new workspaces.