Odor source localization in outdoor building environments through distributed cooperative control of a fleetof UAVs
Meh Jabeen,Qing-Hao Meng,Hui-Rang Hou,Hong-Yue Li
DOI: https://doi.org/10.1016/j.eswa.2024.123332
IF: 8.5
2024-02-03
Expert Systems with Applications
Abstract:Locating the odor source in an outdoor building environment using autonomous vehicles poses a considerable challenge. To address these challenges, researchers have explored the employment of a fleet of unmanned aerial vehicles (UAVs) for odor source localization (OSL). UAVs provide mobility and adaptability, making them a perfect tool for reaching remote or hazardous areas. However, optimizing the flight paths of a fleet of UAVs to efficiently explore and locate odor sources in dynamic outdoor environments is a complex problem. In this study, a novel distributed cooperative control framework was proposed to track and effectively explore the odor source. The framework comprises of two layers: the fleet layer and the control layer. The UAVs in the fleet layer cooperate with one another to share information and set waypoints through the concept of particle swarm optimization (PSO). The waypoints generated in the fleet layer serve as guidance for the UAVs in the control layer, where the perturbation-based extremum seeking control (ESC) is employed to navigate the UAVs towards the odor source. The proposed framework was proved to be stable and ensures steady convergence to the optimum position. Additionally, in order to create a very realistic outdoor environment and utilize a fleet of UAVs for 3-D OSL, the GADEN simulator was integrated with the ROS-Gazebo system to build a simulation platform called Gazebo-GADEN. The performance of the proposed framework was evaluated by comparing it with two benchmark approaches, perturbation-based extremum ESC and standard PSO. On the one hand, the two benchmark approaches may struggle to adapt to the dynamic and uncertain nature of outdoor environments. On the other hand, the proposed framework approach demonstrates remarkable adaptability to the dynamic environment and outperforms the two benchmark approaches in terms of searching time and success rate.
computer science, artificial intelligence,engineering, electrical & electronic,operations research & management science