Hazardous Chemical Source Localisation in Indoor Environments Using Plume-tracing Methods

Abstract

Bio-inspired chemical plume-tracing methods have been applied to mobile robots to detect chemical emissions in the form of plumes and localise the plume sources in various indoor environments. Nevertheless, it has been found from the literature that most of the research has focused on plume tracing in free-stream plumes, such as indoor plumes where the chemical sources are located away from walls. Moreover, most of the experimental and numerical studies regarding the assessment of indoor plume-tracing algorithms have been undertaken in laboratory-scale environments. Since fluid fields and chemical concentration distributions of plumes near walls can be different from those of free-stream plumes, understanding of the performance of existing plume-tracing algorithms in near-wall regions is needed. In addition, the performance of different plume-tracing algorithms in detecting and tracing wall plumes in large-scale indoor environments is still unclear. In this research, a simulation framework combining ANSYS/FLUENT, which is used for simulating fluid fields and chemical concentration distributions of the environment, and MATLAB, with which plume-tracing algorithms are coded, is applied. In general, a plume-tracing algorithm can be divided into three stages: plume sensing, plume tracking and source localisation for analysis and discussion. In the first part of this research, an assessment of the performance of sixteen widely-used plume-tracing algorithms equipped with a concentration-distance obstacle avoidance method, was undertaken in two different scenarios. In one scenario, a single chemical source is located away from the walls in a wind-tunnel-like channel and in the other scenario, the chemical source is located near a wall. It is found that normal casting, surge anemotaxis and constant stepsize together performed the best, when compared with all the other algorithms. Also, the performance of the concentration-distance obstacle avoidance method is unsatisfactory. By applying an along-wall obstacle avoidance method, an algorithm called vallumtaxis, has been proposed and proved to contribute to higher efficiencies for plume tracing especially when searching in wall plumes. The results and discussion of the first part are presented in Chapter 4 of this thesis. In the second part, ten plume-tracing algorithms were tested and compared in four scenarios in a large-scale indoor environment: an underground warehouse. In these four scenarios, the sources are all on walls while their locations are different. The preliminary testing results of five algorithms show that for most failure cases, the robot failed at source localisation stage. Consequently, with different searching strategies at source localisation stage, this research investigated five further algorithms. The results demonstrated that the algorithm with a specially-designed pseudo casting source localisation method is the best approach to localising hazardous plume sources in the underground warehouse given in this research or other similar environments, among all the tested algorithms. The second part of the study is reported in Chapter 5 of this thesis.