Multi-Robot Gas Sources Localization and Mapping Using Adaptive Voronoi-PSO and Bayesian Inference

Abstract

Gas sources localization (GSL) and gas distribution mapping (GDM) are vital when hazardous gas leaks occur, as finding the sources and mapping the contaminant are crucial for containment and prevention. This paper addresses the challenge of simultaneous GSL and GDM through a multi-robot coordination algorithm in a large area. Bayesian inference methods are employed to estimate areas with the highest gas concentration. The robots are directed towards the most probable locations of the highest gas concentration. Even after locating all the gas sources, the robot group focuses on mapping the more contaminated areas. Voronoi tessellation is utilized to partition the working area among the robots. The Particle Swarm Optimization (PSO) method is employed, allowing one robot to share the current highest probability of the gas source’s location to others. The weighting constant of the PSO method is eliminated, enabling GSL and GDM to be performed in an adaptive manner. The proposed method is referred to as “adaptive Voronoi-PSO”. The method is evaluated in both free-space and cluttered environments. Monte Carlo simulation is suitable for free-space environments as it rapidly generates gas distribution. Evaluating the method in cluttered environments is more challenging, requiring complex simulations to generate the gas distribution. Overall, the adaptive Voronoi-PSO method offers more advantages than solely using Voronoi or non-adaptive Voronoi-PSO methods. Its only drawback is observed in cluttered environments. Nevertheless, the adaptive Voronoi-PSO method remains advantageous as it eliminates the need to select a weighting constant, which is a challenging task in practice.

Keywords

• Gas sources localization
• gas distribution mapping
• multi-robot coordination