Configurable Simulation Testing for Autonomous Mobile Multi-Sensor Plume Source Localization

In hazardous situations involving the release of Chemical, Biological, Radiological and Nuclear (CBRN) pollutants, standard emergency procedures include the evacuation of all potentially affected civilian areas in addition to containment of the release. Amidst unpredictable wind and atmospheric conditions, the dispersion of pollutant plumes can vary in speed and direction, introducing severe difficulties in predicting its trajectory (and potential evacuation sites), and ultimately containment efforts. Ongoing developments in estimating rapid contaminant dispersion include the combined use of local meteorological data, satellite imaging, and source localization via autonomous data-guided mobile sensing platforms ranging widely in environmental measurement capabilities and configurations. Ultimately, with a vast number of possible sensor configurations, plume dispersion characteristics, and Source Seeking (SS) guidance algorithms, platform selection for real-world application involves a great deal of uncertainty. This study aims to develop a robust performance-testing simulator to offer a reliable comparison of SS approaches among mobile sensing platforms within a configurable dynamic plume simulation environment. Utilizing Robot Operating System (ROS) and Simulink, a navigable simulation environment was developed for a single mobile robot (Pioneer 3AT) with an optional selection of mounted sensors capable of measuring gas concentration, wind speed and velocity, turbulence, and temperature. 3D time-varying Gaussian plume dispersion models released from a single fixed source were generated using MATLAB and placed within the simulation’s navigable domain. Varying atmospheric conditions and dispersion characteristics were established for use on a case-by-case basis. For selected sensor configurations, individual SS approaches of increasing complexity (chemotaxis, anemotaxis, combined approaches) containing a range of associated source localization algorithms were tested with several dispersion model cases. The shortest amount of time required for each algorithm to locate the fixed source within a distance of one meter was used as a performance measure, with the distance between the vehicle and source location throughout navigation (SS error) offering an additional method of comparison. The resulting time and SS error for each localization algorithm are then compared with the results obtained from current simulation and experimental studies involving autonomous ground-based wheeled vehicle source localization. By introducing adjustable dynamic dispersion models to a configurable mobile sensing robot simulation environment, comparable evaluation of methods aimed at observing and forecasting hazardous environmental conditions can be accomplished prior to any applicable implementation.