Bayesian Search Algorithm for Multi-Drone Deployment

Searching a broad area for a source of unknown composition or strength comes with inherent challenges; it is often difficult to distinguish a radioactive source over naturally occurring background radiation if the counts from the source are comparable to the background field strength, there is no directional information returned from the detector, and time is critical. Bayesian statistics makes this problem more feasible. By treating the search area as a probability space, incoming data from the detector can be used to update the probability space as to the most probable location of the illicit or lost source. The result, known as the posterior probability, is improved by one of two methods; a more accurate initial estimate of the source location or more data to improve the confidence of the posterior probability distribution. More often than not there is no prior information about the location of a source within a given space, so it is best to treat the search space as an equally distributed probability initially. Thus this leaves the option of acquiring more data. Sending a swarm of drones into a search space drastically increases the amount of data that can be collected in a given amount of time; improving the ability to detect an anomaly as well as improving the speed at which a source can be located. The novel work being done on this project aims to develop an algorithm capable of employing multiple drones into a search space, and using Bayesian inference techniques efficiently locate a radioactive, chemical, or biological source.