SOURCE LOCALIZATION AND MAPPING OF CHEMICAL GAS CONTAMINATED AREA WITH MULTI UAV

• Main Author: Aris Prabowo, Yaqub
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77343
• Institution: Institut Teknologi Bandung

In the field of CBRN (Chemical, Biological, Radiological, and Nuclear), hazardous substances can be caused by industrial accidents, acts of terrorism, or natural disasters, resulting in mass and significant damage to an area. Currently, most industries or certain strategic areas use static CBRN sensors to obtain maps of hazardous substance distribution. In reality, statically mounted sensors can be damaged due to explosions, cannot reach all areas, and are expensive because many sensors need to be installed. Therefore, an unmanned or olfactory robot is needed. An olfactory robot is a robot equipped with sensors that can detect CBRN substances. In this study, the problem is limited to map chemical substances or gases only. To map chemical gases in a wide outdoor environment, a robot with good maneuvering, scalability, and flexibility capabilities, such as a rotary-wing type Unmanned Aerial Vehicle (UAV), is required. A UAV equipped with a chemical gas sensor can be commanded to move based on chemical gas readings in contaminated areas. Multi-UAVs are used to speed up the source localization and distribution mapping mission in a large environment. A research of how to control one or several UAVs for chemical gas source localization and distribution mapping is addressed in this dissertation. It is assumed that there is a warning due to a gas leak in a certain area, either from an alarm installed or from reports from the surrounding community. After receiving the warning, the UAV will be flown to carry out reconnaissance in the area. The purpose of the reconnaissance is to locate the source of the gas and estimate the gas distribution map. The predicted location of the source of the gas is useful for the neutralization process, stopping the exposed gas leak while the estimated gas distribution map will be useful for assisting the evacuation and mitigation process. Because the flight time is limited, multi UAVs cannot explore the entire area in the environment. Therefore, each UAV must effectively carry out exploitation and exploration in the area defined as the chemical gas-exposed area. In this case, exploitation means the UAV heading towards the source of the gas, and exploration means the UAV heading towards another area that is not the source of the gas. iv Good exploration means the UAV covers more interesting areas, that is, areas with high concentrations, even if they are not near the source of the gas. To enable the UAV to estiamate the location of gas source or the area with high gas concentration, it was decided to utilize the Bayesian inference method. In the Bayesian inference method, a way to estimate the likelihood probability of where the gas source is or where the high concentration of gas is located is needed. This estimation is quite difficult due to the complex movement of gas particles, influenced by gravity, wind movement, and diffusion. The gas movement is even more complex in outdoor environments with many buildings and objects. In this research, it is assumed that the core building models have been known, so some wind movement profiles can be simulated first. Some of these wind movement profiles are used to simulate the movement of gas particles. Each simulation of gas movement with the gas source placed in different areas can be represented as a probability density function or PDF that can be used to estimate the likelihood in the Bayesian inference method. The anemotaxis method is also used, integrated with the Bayesian inference method for more efficient performance. In the UAV coordination scheme, the Voronoi method is utilized to decompose the area for each UAV. This decomposition is dynamic so that the UAV always has its own exploration area. The determination of exploration location is also modified, inspired by the Particle Swarm Optimization (PSO) method so that multi-gas source localization and gas distribution mapping can be done more efficiently. Testing is carried out gradually, starting with the use of mobile robots in indoor environments, one UAV in outdoor environments both simulation and real-world experiments, and multi UAVs in outdoor environments that are simulated. Efficiency and effectiveness of the proposed methods are measured with several metrics: (1) time finding the gas source, (2) area of contaminated region, (3) Negative Log Predictive Density (NLPD) representing the map accuracy. From the evaluation analysis, it can be concluded that the proposed methods show their efficiency and effectiveness in gas sources localization and distribution mapping.