Yash Chitgopekar


Designing Markov Chain Algorithms for Enhanced Robotic Surveillance

Modern computational power and the newfound prevalence of data have rendered standard, static surveillance systems vulnerable and incapable of providing protection to vital resources, institutions, and citizens. In contrast to the deterministic surveillance strategies used today, stochastic approaches to autonomous surveillance retain the advantage of being random and unpredictable. The development of robotic control algorithms that incorporate randomness is therefore useful to the advancement of robotic surveillance as a whole. By building Markov chain models that serve as a specific decision-making policy for autonomous robots and testing their performance through Matlab simulations over various topological networks and intruder models, we discover that the maxentropic Markov chain achieves a relatively high detection rate for intelligent intruders with short attack durations. This result is important for the realistic scenario in which an intruder plans its attacks according to the robotic agent’s prior actions to minimize the possibility of detection.

UC Santa Barbara Center for Science and Engineering Partnerships UCSB California NanoSystems Institute UC Santa Barbara’s Parents Fund Campaign for UC Santa Barbara