Experimental Study On Multiple Odor Sources Mapping By A Mobile Robot In Time-Varying Airflow Environment

This paper addresses the experimental study on mapping multiple odor sources using the proposed algorithm based on Dempster-Shafer (D-S) theory in outdoor environments where the airflow almost always be strong and change with time both in direction and speed. In the proposed algorithm, by using the time series of the measurements from an anemometer and twelve gas sensors, the robot estimates the air-mass path to compute the belief mass function, and iteratively updates a grid map with D-S inference at each time step, mapping the distribution of two odor sources when the robot is cruising in the given search area. Here the air-mass path is defined as the historical trajectory most likely taken by the air mass contacting with the gas sensor mounted on the mobile robot, and an air mass might contain odor molecules, or just be a mass of clean air. In the grid map, each cell has two possible states, being occupied by an odor source, or not. In order to cut the calculation of the proposed algorithm, only the cells covered by the estimated air-mass path are updated. Experiments were carried out in outdoor nature environments, and the results illustrate that the proposed algorithm based on D-S theory can be applied to map multiple odor sources online.