4CNet: A Diffusion Approach to Map Prediction for Decentralized Multi-Robot Exploration

TL;DR

This paper introduces 4CNet, a deep learning architecture that enhances decentralized multi-robot exploration by improving map prediction accuracy and efficiency under resource constraints, using a novel confidence-aware contrastive learning approach.

Contribution

The paper presents 4CNet, a novel deep learning model that integrates conditional consistency, contrastive pretraining, and confidence estimation for improved map prediction in multi-robot exploration.

Findings

4CNet-E achieves higher map prediction accuracy than state-of-the-art methods.

4CNet-E significantly improves area coverage in diverse environments.

Hardware tests confirm the model's effectiveness in real outdoor and indoor settings.

Abstract

Mobile robots in unknown cluttered environments with irregularly shaped obstacles often face energy and communication challenges which directly affect their ability to explore these environments. In this paper, we introduce a novel deep learning architecture, Confidence-Aware Contrastive Conditional Consistency Model (4CNet), for robot map prediction during decentralized, resource-limited multi-robot exploration. 4CNet uniquely incorporates: 1) a conditional consistency model for map prediction in unstructured unknown regions, 2) a contrastive map-trajectory pretraining framework for a trajectory encoder that extracts spatial information from the trajectories of nearby robots during map prediction, and 3) a confidence network to measure the uncertainty of map prediction for effective exploration under resource constraints. We incorporate 4CNet within our proposed robot exploration with map prediction architecture, 4CNet-E. We then conduct extensive comparison studies with 4CNet-E and state-of-the-art heuristic and learning methods to investigate both map prediction and exploration performance in environments consisting of irregularly shaped obstacles and uneven terrain. Results showed that 4CNet-E obtained statistically significant higher prediction accuracy and area coverage with varying environment sizes, number of robots, energy budgets, and communication limitations when compared to database and learning-based methods. Hardware experiments were performed and validated the applicability and generalizability of 4CNet-E in both unstructured indoor and real natural outdoor environments.