Robot odor source localization in indoor environments based on gradient adaptive extremum seeking search

Abstract

When toxic or harmful chemicals or pollutants leak, it is of great significance to determine the leakage source(s) in a timely and autonomous manner to reduce casualties and property losses. This study focuses on the problem of robot-based odor source localization (OSL) in unknown indoor environments from a control perspective. A model-free gradient adaptive extremum seeking control (GA-ESC) algorithm was proposed to improve the searching efficiency and success rate of OSL. The GA-ESC algorithm plans the OSL through a controller, which estimates the gradient of the odor plume concentration, and guides the robot to approach the odor source according to the estimated gradient. A novel three-point gradient estimation method was proposed to obtain a more stable odor field gradient based on two historical points and one current point, where an adaptive feedback gain was applied to link the estimated gradient and the output control quantities. Subsequently, the perturbation amplitude adjustment (PAA) strategy was introduced to enhance the global searching ability. When compared to the traditional extremum seeking control (ESC), our algorithm outperforms it in terms of success rate and average searching time. Moreover, the performance of the proposed algorithm was validated through simulations by utilizing three different dispersion models. The real-robot experiments were also carried out in indoor environments. The results demonstrate the significance of the proposed control-based OSL algorithm in unknown indoor environments.