Chemical Source Localization Using Virtual Physics Based Robots

In recent years, with the increasing accidental release of hazardous chemicals and the increased risk of biochemical terrorism, researchers have begun investigating mobile robots system equipped with electrical gas/odor sensors for quick location and containment of hazardous chemical sources to reduce casualties. Chemical source localization using mobile robots system is representative of a wide variety of applications. For instance, detect toxic/harmful gas leakage, fight against antiterrorist attacks, rescue in a building on fire, and search for explosives and demining operations.Most prior work related to chemical source localization was devoted to using either single or multiple robot systems for seeking, detecting, and tracking of a single chemical source location. However, very little attention has been given to the equally important problem of localization of multiple chemical sources. This problem has practical relevance to situations where it is imperative to simultaneously identify and locate all the chemical sources before they can cause a great loss to the environment and people in the vicinity. The above problem presents several challenges such as multiplicity and cross interference of different sources, need for simultaneous localization of all sources.To begin with setting up simulation environment, this paper focuses on a single chemical source and multiple chemical sources localization using virtual physics and species based robots.Firstly, considering the drawbacks of the current plume model to simulate indoor ventilation environments and outdoor ventilation environments with multiple chemical sources, a two-dimension indoor turbulent plume model and two two-dimension outdoor turbulent plume models with two and five chemical sources used to verify the chemical source localization algorithm are built by the FLUENT software. At the same time, a simulated robot model is given. So, simulation environment based MATLAB software used to visualize the process of chemical source localization is set up.Secondly, for the problem of single chemical source localization in the indoor arena, a multi-robot cooperation strategy with virtual physics force, this includes three kinds of effort: structure formation force, goal force, and obstacle avoidant force, is proposed. Three kinds of plume tracing algorithms: fluxotaxis, chemotaxis and anemotaxis are given by constructing the goal force. With the balance of the forces acted on the robots, a chemical source declaration is brought forward by computing odor mass flux. Simulation comparison experiments in the two dynamic indoor plume models(without obstacle and with obstacles) using the proposed three kinds of plume tracing algorithms are carried out respectively and the comparative result is illustrated. At the same time, the influences of the robot’s number, the initial position of the robots and the varied wind direction/ speed frequency and chemical release frequency to the search performance are discussed.Thirdly, with the problem of two chemical sources localization in outdoor ventilation environments, parallel searches by two groups of robots are used to make searches faster. In order to avoid re-finding the same source, a multi-robot cooperation strategy with virtual physics force, repulsion force and rotary force, is proposed. Simulation experiments in the two dynamic outdoor plume models(without obstacle and with obstacles) discuss the influence of the varied frequencies of wind direction/ speed and methane release with different initial positions of multiple groups to the search performance. Simulation comparison experiments using three kinds of plume tracing algorithms: chemotaxis, anemotaxis and fluxotaxis are carried out respectively and the comparative result about three plume tracing algorithms is illustrated.Fourthly, in concern with five chemical sources localization, a release mechanism is introduced into the multi-robot cooperation strategy to realize the robots continue searching for other sources once a source has been found. The release mechanism includes two parts: taboo domain setting and rotary force setting with which the group of robots can escape from the taboo domain and avoid a source to be re-localized. First, simulation experiments in the two chemical sources plume models(without obstacle and with obstacles) are re-done by one group of robots using chemotaxis in order to improve the simulation result by two groups of robots. Second, simulation experiments using two groups of robots in the five chemical sources plume model discuss the influence of the radius of the taboo domain, the varied frequencies of wind direction/ speed and methane release to the search performance. Finally, simulation comparison experiments using three kinds of plume tracing algorithms: chemotaxis, anemotaxis and fluxotaxis by one group and two groups of robots are carried out respectively and the comparative result is illustrated.Finally, for the needs of multiple chemical sources localization in the indoor arena, a multi-robot cooperation strategy based on a modified glowworm swarm optimization(M-GSO) is proposed. This strategy includes global random search of self-exploration, local search based on GSO algorithm and chemical source declaration. And forbidden area setting is also introduced into the iteration process to achieve localization for multiple chemical sources. This mechanism can ensure robots to continue searching for the other sources once a source has been found and ensure that other robots would not re-locate this chemical source. The odor plume is found by global random search of self-exploration which includes spiral search and wave-shape search. Spiral search enables the robots perform an extensive search, while the wave search has a strong direction and can make the robot probe deeper into the search field. Plume tracing algorithm with the modified glowworm swarm optimization(M-GSO), in which variable step length is used, is proposed. The dynamic decreasing step length is used when all the robots are initially placed in the edge of the search space, while the random step length following a uniform distribution or a norm distribution is used when the robots are initialized with a random position in the search space. A three-step chemical source declaration algorithm is also brought forward. Simulation results show that the proposed M-GSO can effectively enable the robot system to search and locate all the chemical sources existed in the indoor environment quickly and accurately.There are 118 Figures, 32 Tables and 167 References in this paper.