GP-Frontier for Local Mapless Navigation

TL;DR

This paper introduces the GP-Frontier, a novel local navigation method for robots that uses Gaussian Process uncertainty to navigate safely without a map, suitable for unknown environments.

Contribution

The paper presents the Gaussian Process Frontier, a new frontier concept leveraging sparse Gaussian Processes for mapless robot navigation using only local sensing.

Findings

Successfully navigates in unknown environments

Reduces collision risk by moving in open spaces

Operates without a pre-built map or path planner

Abstract

We propose a new frontier concept called the Gaussian Process Frontier (GP-Frontier) that can be used to locally navigate a robot towards a goal without building a map. The GP-Frontier is built on the uncertainty assessment of an efficient variant of sparse Gaussian Process. Based only on local ranging sensing measurement, the GP-Frontier can be used for navigation in both known and unknown environments. The proposed method is validated through intensive evaluations, and the results show that the GP-Frontier can navigate the robot in a safe and persistent way, i.e., the robot moves in the most open space (thus reducing the risk of collision) without relying on a map or a path planner.