Multi-Robot Odor-Source Localization In Turbulence Dominated Airflow Environments

The sense of smell has been widely used in biological world. Animals use odors to exchange information, find mates, evade predators, search for food and so on. Inspired by the biological olfaction, in the early 1990s researchers started to build mobile robots with onboard gas/odor sensors achieving the so-called active olfaction, which include the functions such as odor source localization, odor distribution mapping and odor path guidance.The research on active olfaction is related to the research fields such as mobile robot navigation and control, sensing and information processing, bionics, computation intelligence and hydrodynamics. It is expected that robot active olfaction will play more and more roles in the application areas like judging toxic/harmful gas leakage location, checking contraband, searching for survivors in collapsed buildings, antiterrorist attacks and so on.This dissertation focuses on the multi-robot based odor source localization in turbulence dominated airflow environments. The achievement can be concluded as follows:First, in view of the drawbacks using the current plume model to simulate indoor ventilation environments, a two-dimension indoor turbulent plume model used for verifying the odor source localization algorithm is built by combining the theory of computational fluid dynamics and partial idea of current plume model. A second-order odor sensor model which considers the real response and recovery time of odor sensors is set up.Second, the algorithms for multi-robot based plume finding, plume tracking as well as odor source declaration are put forward. The odor plume is found by multiple robots via divergent search and gradual coverage. Two plume tracking algorithms, Upwind surge combined with Modified Ant Colony Optimization (U-MACO) as well as Probability-fitness-function Particle Swarm Optimization (P-PSO), are proposed. Both the tracking algorithms can use the wind and odor information effectively and can avoid falling into local extrema by overcoming the influence of complex characteristics of dynamic plumes. A three-step odor source declaration algorithm is brought forward by combining heuristic thoughts and odor mass throughput calculation. Third, simulation experiments in both the outdoor large-scale and indoor small-scale plume models demonstrate that the proposed multi-robot odor source localization algorithms are superior to the previous particle swarm optimization based algorithm. The proposed algorithms are also validated via real robots experiments in indoor ventilated environments.Finally, the characteristics of Exploration and Exploitation (E-E) in multi-robot based odor source localization are analyzed. The index and classification method for balancing the E-E are proposed. The MACO and PSO are combined with the proposed two E-E balance modes, and simulation experiments show that the asynchronous mode is better than that of synchronous mode.