Experimental Research On The Multi-robot Plume Tracing Algorithms

Odor source localization (OSL) is the behavior of finding the location of a volatile chemical source in the environment. It is vital to many creatures in nature. They use odor information to search for food, find preys, escape their predator, locate mates, and so on. Developing robots with such skills could be used to gain deeper understanding of the above behavior of creatures. The potential applications of the OSL include: finding the emitter source of toxic and hazardous gases, detecting contrabands such as drugs, searching for the survivors in earthquakes, typhoons, and mine disasters, and fighting against terrorist attacks.The research on the robot based OSL could trace back to the early 1990s. More and more researchers and research institutions have paid attention to the OSL due to its wide application prospects. Compared to the single-robot based OSL, less research on the multi-robot OSL has been reported.This thesis focuses on the multi-robot OSL in indoor natural wind environments. The research work can be concluded as follows:Firstly, the state-of-art of the OSL using both single and multiple robots is reviewed. Typical algorithms for plume finding, plume tracing and odor source declaration are summarized.Secondly, the localization of multiple mobile robots is realized by using global vision combined with dead-reckoning. The multi-robot obstacle-avoiding capability based on traffic regulations is acquired under the subsumption architecture. The gas sensors mounted on the robots are calibrated using the simplified formula.Thirdly, experiments for multi-robot plume tracing using the improved ant colony optimization (ACO) algorithm is carried out in indoor natural wind environments. The spiral surge (SS) algorithm is also conducted for comparison. Experimental results show that the improved ACO performs better than the SS in terms of search efficiency and robustness. Based on the observation of the trials, it is found that the OSL performance can be improved through decreasing interference of the behaviors such as obstacle avoidance. Several feasible solutions are proposed.