Scalable Gas Sensing, Mapping, and Path Planning via Decentralized Hilbert Maps.

Area of Science:

• Robotics and Autonomous Systems
• Environmental Sensing
• Distributed Computing

Background:

• Accurate gas distribution mapping (GDM) is crucial for environmental monitoring and safety.
• Existing GDM methods often rely on centralized systems, limiting scalability and robustness.
• Information-driven path planning is essential for efficient exploration by distributed sensing systems.

Purpose of the Study:

• To develop a decentralized approach for gas distribution mapping (GDM) using Hilbert maps.
• To create an information fusion method for merging data from multiple robots efficiently.
• To devise novel entropy-based information-driven path-planning algorithms for distributed sensing systems.

Main Methods:

• Utilized Hilbert maps with kernel logistic regression for probabilistic GDM as a classification task.
• Developed a novel Hilbert map information fusion technique for decentralized data merging.
• Implemented a communication strategy for collaborative GDM computation among multiple robots.
• Introduced new entropy-based information-driven path-planning methods and compared them with Particle Swarm Optimization (PSO) and Random Walks (RW).

Main Results:

• The proposed decentralized GDM approach effectively merges information from individual robot maps with limited communication.
• The novel information-driven path-planning methods significantly outperformed PSO and RW in simulated environments.
• The system demonstrated real-time avoidance of mutual collisions among robots during path planning.

Conclusions:

• Decentralized GDM using Hilbert maps and information fusion enables efficient large-scale distributed sensing.
• Entropy-based information-driven path planning offers superior performance for exploration and navigation in multi-robot systems.
• The developed methods provide a robust and scalable solution for real-time GDM and autonomous navigation.