Improving Visual Place Recognition Based Robot Navigation By Verifying Localization Estimates

TL;DR

This research introduces a novel Multi-Layer Perceptron (MLP) integrity monitor which demonstrates improved performance and generalizability, removing per-environment training and reducing manual tuning requirements.

Abstract

Visual Place Recognition (VPR) systems often have imperfect performance, affecting the `integrity' of position estimates and subsequent robot navigation decisions. Previously, SVM classifiers have been used to monitor VPR integrity. This research introduces a novel Multi-Layer Perceptron (MLP) integrity monitor which demonstrates improved performance and generalizability, removing per-environment training and reducing manual tuning requirements. We test our proposed system in extensive real-world experiments, presenting two real-time integrity-based VPR verification methods: a single-query rejection method for robot navigation to a goal zone (Experiment 1); and a history-of-queries method that takes a best, verified, match from its recent trajectory and uses an odometer to extrapolate a current position estimate (Experiment 2). Noteworthy results for Experiment 1 include a decrease in aggregate mean along-track goal error from ~9.8m to ~3.1m, and an increase in the aggregate rate of successful mission completion from ~41% to ~55%. Experiment 2 showed a decrease in aggregate mean along-track localization error from ~2.0m to ~0.5m, and an increase in the aggregate localization precision from ~97% to ~99%. Overall, our results demonstrate the practical usefulness of a VPR integrity monitor in real-world robotics to improve VPR localization and consequent navigation performance.

Figures (9)

• Figure 1: Overview of our system, demonstrating how our addition of a predictive verification system (a Multi-Layer Perceptron (MLP) network) results in safer navigation. See Figure \ref{['fig:MLPschem']} for more.
• Figure 2: An overview of how inputs for the MLP integrity monitor are extracted. For a given query, we use four vectors formed from the output of the VPR process. The vectors pass through a statistical feature extractor and are then concatenated. These steps guarantee the size of the input is suitable for the MLP network, which accepts a 192-dimensional vector.
• Figure 3: HoQ method: we rank a recent history of VPR matches by their match distance (1), reduce to verified matches (2), then extrapolate from the best, verified match to estimated position using the odometer difference (3).
• Figure 4: In Experiment 1, the robot moves towards the end-goal, and makes a navigation decision for each verified VPR match.
• Figure 5: Photo bank of sample dataset images, grouped into columns per environment. Adversities include anomalous obstacles, dynamic objects, lighting changes, camera occlusion, glare, and time-of-day changes.