Radiation detection using mobile sensor networks
Date
2016
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
This dissertation addresses a fixed-time interval decision making problem from an optimal control perspective, in the context of deciding the radioactive nature of a single target in transit using multiple mobile sensors. The mobility of sensors facilitates detecting extreme weak sources that may otherwise slip through stationary radiation sensor networks. However challenging problems also arise in terms of searching for the optimal way to utilize such mobility.
Based on existing work, the decision on target's radioactive nature is made through a Likelihood Ratio Test (LRT), whose probability of making mistakes are upper bounded by analytic expressions related to both LRT thresholds and trajectories of mobile sensors. This dissertation proposes a threshold selection process for the LRT based on the constraint on its false alarm rate and solves for the optimal sensor trajectories that would minimize the upper bounds on the probability of missing detection.
Under simplifying assumptions on the motion and geometry of the source, the sensors, and the surrounding environment, the optimal control problem admits an intuitive, analytic closed-form solution. The intuition derived from this analytic solution supports the development of a motion control law that steers (suboptimally) the sensors to a given neighborhood of the suspected source, while navigating among stationary obstacles in their environment. This motion controller closes the loop at the acceleration level of a heterogeneous collection of sensor platforms.
This dissertation detailed a robot control system developed for conducting the radiation detection experiment using physical platforms, which is capable of controlling multiple robots simultaneously. Experimental studies with these robots corroborate the theoretical convergence results of the proposed navigation controller. The detection of weak radioactive source (Vaseline beads) is achieved with these sensor platforms.
The limitation of this work is that it relies on some strong localization assumptions as the optimal strategy requires the position feedback of both the robots and the target, which might be difficult to obtain in some scenarios. One extension discussed in this dissertation is to study the possibility of achieving convergence to signal extremum without global localization in cluttered environment. In this dissertation, the unavoidable collisions resulted from the lack of localization is modeled as a Markov process and the effect of these collisions on the probability of the robot successfully converging to the signal extremum is studied. Future studies on the reachability problem of the underlying stochastic hybrid system could lead to better usage of the multiple sensor platforms to survey uncertain areas.