Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Exploration Efficiency using Uncertainty-Aware Information Prediction

Seunghwan Kim, Heejung Shin, Gaeun Yim, Changseung Kim, Hyondong Oh

TL;DR

The paper addresses efficient autonomous exploration in unknown environments by incorporating uncertainty-aware neural network predictions of occupancy grids into information-guided planning. It trains a Bayesian neural network to predict local OGMs near frontiers and uses MC sampling to quantify predictive uncertainty, integrating this into MI-based metrics including Uniform FSMI with Map Prediction. Four map-prediction–based information metrics are proposed and validated through frontier-exploration simulations, showing shorter exploration times and improved robustness when uncertainty is accounted for. The work advances reliable, uncertainty-aware exploration with practical extensions toward real-world deployment and handling of aleatoric uncertainty.

Abstract

Autonomous exploration is a crucial aspect of robotics, enabling robots to explore unknown environments and generate maps without prior knowledge. This paper proposes a method to enhance exploration efficiency by integrating neural network-based occupancy grid map prediction with uncertainty-aware Bayesian neural network. Uncertainty from neural network-based occupancy grid map prediction is probabilistically integrated into mutual information for exploration. To demonstrate the effectiveness of the proposed method, we conducted comparative simulations within a frontier exploration framework in a realistic simulator environment against various information metrics. The proposed method showed superior performance in terms of exploration efficiency.

Enhancing Exploration Efficiency using Uncertainty-Aware Information Prediction

TL;DR

The paper addresses efficient autonomous exploration in unknown environments by incorporating uncertainty-aware neural network predictions of occupancy grids into information-guided planning. It trains a Bayesian neural network to predict local OGMs near frontiers and uses MC sampling to quantify predictive uncertainty, integrating this into MI-based metrics including Uniform FSMI with Map Prediction. Four map-prediction–based information metrics are proposed and validated through frontier-exploration simulations, showing shorter exploration times and improved robustness when uncertainty is accounted for. The work advances reliable, uncertainty-aware exploration with practical extensions toward real-world deployment and handling of aleatoric uncertainty.

Abstract

Autonomous exploration is a crucial aspect of robotics, enabling robots to explore unknown environments and generate maps without prior knowledge. This paper proposes a method to enhance exploration efficiency by integrating neural network-based occupancy grid map prediction with uncertainty-aware Bayesian neural network. Uncertainty from neural network-based occupancy grid map prediction is probabilistically integrated into mutual information for exploration. To demonstrate the effectiveness of the proposed method, we conducted comparative simulations within a frontier exploration framework in a realistic simulator environment against various information metrics. The proposed method showed superior performance in terms of exploration efficiency.

Paper Structure

This paper contains 20 sections, 22 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Occupancy Grid Mapping
  4. Frontier Exploration
  5. Frontier Detection
  6. Viewpoint Generation
  7. Select the Best Frontier
  8. Volumetric Gain
  9. Uniform FSMI
  10. Information Prediction
  11. Occupancy Grid Map Prediction Using Bayesian Neural Network
  12. Information Metric with Map Prediction
  13. Volumetric Gain with Map Prediction
  14. Variance of Prediction Map
  15. Variance of Prediction Map with Predicted Raycasting
  16. ...and 5 more sections

Figures (5)

  • Figure 1: Sample of the viewpoint generation.
  • Figure 2: Occupancy grid map prediction at the $i$-th frontier using a deep neural network.
  • Figure 3: Illustrative results for each information metric. The red arrow indicates the position and direction of the sensor. In the uniform FSMI and variance of prediction map, colors closer to yellow represent higher information, while colors closer to purple indicate lower information.
  • Figure 4: Simulation environments.
  • Figure 5: Autonomous exploration simulation results. Each plot represents the average of 10 trials.