Design and Performance Evaluation of a Particle Filter-based Algorithm for Smoke Plume Path Tracking

We handle the problem of tracing the particulate matter plumes from a smoke source with a mobile robot equipped with a laser particle counter-based smoke sensor in this paper. The proposed method is developed from a particle filter-based chemical odor source localization algorithm and makes a few modifications to adapt it to smoke plumes. Based on the differences between particulate smoke plumes and chemical plumes in terms of distribution, we build a general smoke plume path model used to update particle weight in the particle filter-based algorithm. We also let the algorithm change its observation threshold adaptively according to the robot’s history measurement values. The algorithm is evaluated in two different simulation environments, with an uniform airflow field and a space-variant airflow field respectively. The proposed algorithm shows good performance in both environments regardless of the location of the starting position of the robot.