Abstract
The problem is autonomous coordinated search by an interconnected group of moving robots for the purpose of finding and localising a source of hazardous emissions (e.g., gas and particles). Dispersion of the emitted substance is assumed to be affected by turbulence, resulting in the absence of concentration gradients. The chapter proposes a search strategy that operates in a completely decentralised manner, as long as the communication network of the moving robots forms a connected graph. By decentralised operation, we mean that each moving robot is reasoning (i.e., estimating the source location and making decisions on robot motion) locally. Coordination of the group is achieved by consensus via communication with the neighbours only, in a manner which does not require global knowledge of the communication network topology.
Introduction
Locating the source of a chemical leak or biological hazard is a high-stakes "needle in a haystack" problem. In calm conditions, robots can simply follow a concentration gradient. However, in real-world scenarios, turbulence shreds these smooth gradients into a chaotic mess of disconnected "patches" of gas. While "infotaxis"—a strategy where robots move to maximize information gain—works well for single robots, scaling this to a team has traditionally required a central controller. Centralized systems are fragile and do not scale well. This research introduces a fully decentralized, scalable, and robust framework for a swarm of robots to locate hazardous sources in turbulent environments.
Methodology
The proposed framework integrates three core components to enable decentralized search:
- Infotaxis (Information-Theoretic Search): Robots prioritize minimizing the entropy (uncertainty) of their internal map regarding the source location, rather than simply seeking the highest concentration.
- Rao-Blackwellised Particle Filter (RBPF): Each robot maintains an RBPF to estimate the source's 2D position and strength. By analytically calculating the source's strength using a Gamma distribution, the RBPF reduces computational overhead, making it suitable for onboard hardware.
- Average Consensus Protocol: This protocol allows the swarm to coordinate without a central leader. Robots share local measurements with immediate neighbors, iteratively reaching a consensus on the optimal next move and maintaining formation.
Experimental Validation and Results
The algorithm was validated using a real-world experimental dataset from COANDA Research & Development Corporation, involving fluorescent dye released into a large recirculating water channel. The researchers performed 200 Monte Carlo simulations with a formation of $N=7$ robots in an upscaled $500 \times 500$ search area.
Key results and insights include:
- Efficiency and Reliability: The mean search time was approximately 2525 arbitrary units (a.u.), with consistent performance despite the chaotic nature of the environment.
- Connectivity: The system only requires the communication graph to remain connected; global network topology knowledge is unnecessary.
- Scalability: Computational and communication loads per robot remain constant as the swarm size increases, as each unit only interacts with its immediate neighbors.
- Robustness: The lack of a leader-follower hierarchy ensures that the failure of a single robot does not compromise the mission.
Conclusion and Future Work
The study demonstrates a robust method for decentralized source localization in turbulent plumes. The current model assumes known environmental parameters like wind speed and direction. Future work will focus on the algorithm's sensitivity to these parameters and developing systems capable of estimating environmental constants on the fly in complex, real-world emergency scenarios.
Summary updated