Information-Theoretic Autonomous Source Search and Estimation of Mobile Sensors

Estimation of a source term, including the origin and release rate, for reconstructing a hazardous chemical, biological, or radiological substance dispersion event in the atmosphere is very important for public safety. The increase in the potential danger of hazardous substances leakage accidents and the threat of malicious acts in random places makes the estimation of the source term difficult using traditional systems such as pre-installed ground sensors in specific areas or ground vehicles. Unmanned aerial vehicles (UAVs) can be considered as an alternative solution for estimating the source term because they can be deployed to any arbitrary place and rapidly cover relatively larger areas compared with ground-based systems. This chapter introduces autonomous source search and estimation strategies for UAVs. Bayesian inference-based estimation approaches that can accurately estimate the source term in turbulent and noisy environments are presented using domain knowledge such as the plume dispersion and sensor models. In particular, since the estimation problem is highly nonlinear and non-Gaussian, the sequential Monte Carlo method (i.e., particle filter) Besides, various information-theoretic decision-making strategies are introduced using different information measures to determine the most informative sampling point at each time step using different information measures. To use the interaction and information sharing among multiple agents at best, cooperation and sensor fusion strategies are also discussed. Finally, comprehensive numerical simulations and flight experiments are presented to validate and compare the performance of the proposed strategies.