Optimizing Gaze Direction in a Visual Navigation Task

TL;DR

This paper presents a method for optimizing gaze direction in visual navigation by modeling active sensing as a POMDP with a mutual information reward, dynamically switching camera configurations to improve navigation.

Contribution

It introduces a novel active sensing approach that dynamically switches camera configurations in a POMDP framework for improved visual navigation.

Findings

Enhanced navigation performance in simulations and real robot experiments.

Dynamic camera configuration switching improves information gathering.

Mutual information-based reward effectively guides gaze control.

Abstract

Navigation in an unknown environment consists of multiple separable subtasks, such as collecting information about the surroundings and navigating to the current goal. In the case of pure visual navigation, all these subtasks need to utilize the same vision system, and therefore a way to optimally control the direction of focus is needed. We present a case study, where we model the active sensing problem of directing the gaze of a mobile robot with three machine vision cameras as a partially observable Markov decision process (POMDP) using a mutual information (MI) based reward function. The key aspect of the solution is that the cameras are dynamically used either in monocular or stereo configuration. The benefits of using the proposed active sensing implementation are demonstrated with simulations and experiments on a real robot.