Feature Based Potential Field for Low-level Active Visual Navigation

TL;DR

This paper introduces a feature-based potential field method for active visual navigation, enabling a vehicle to improve localization by favoring feature-rich areas while moving towards a goal.

Contribution

It presents a novel potential field approach that associates image features with potential energies to guide active visual navigation.

Findings

Effective in guiding quadrotor towards goals

Favors feature-rich areas for better localization

Demonstrates improved navigation performance

Abstract

This paper proposes a novel solution for improving visual localization in an active fashion. The solution, based on artificial potential field, associates each feature in the current image frame with an attractive or neutral potential energy. The resultant action drives the vehicle towards the goal, while still favoring feature rich areas. Experimental results with a mini quadrotor equipped with a downward looking camera assess the performance of the proposed method.