PERIMETER DEFENSE WITH MULTI-ROBOT SYSTEMS

Loading...
Thumbnail Image
Degree type
Doctor of Philosophy (PhD)
Graduate group
Computer and Information Science
Discipline
Electrical Engineering
Computer Sciences
Electrical Engineering
Subject
control
game theory
graph theory
perimeter defense
robotics
Funder
Grant number
License
Copyright date
2024
Distributor
Related resources
Author
Chen, Austin, Ku
Contributor
Abstract

Perimeter defense is a key challenge in many real-world scenarios such as airport security, convoy protection, and self-driving cars. To successfully achieve perimeter defense, a defender must solve several fundamental problems such as target tracking, resource allocation, and pursuit evasion. It is also crucial to consider the capabilities of defenders when creating perimeter defense strategies; a control policy intended for fast agents may perform poorly when used with slow agents. Existing literature has divided a defender's overall fitness into physical capability, which determines how well it may deter intruders, and informational capability, which encompasses knowledge of the intruder's location and actions. Perimeter defense strategies have been investigated across a broad range of physical and informational capabilities and can change substantially based upon the team's relative strengths. This dissertation expands the perimeter defense field by considering previously unexplored regions of defender team capability. In contrast to previous works which only explore scenarios with limited information, we investigate a defender that must contend with adversarially manipulated information. The adversary in this situation leverages its informational advantage to deceive the defender and achieve higher task performance. In addition, we consider a new axis of multi-robot capability which describes the team sizes of both the defender and intruder. This direction is especially important to consider given the recent advances in multi-robot systems. We present control policies across a range of physical and informational capabilities with the added assumption of arbitrarily large teams. These approaches are computationally efficient and demonstrate an improvement in empirical performance over few-agent extensions from previous works. The presented contributions show the importance of using the correct perimeter defense algorithm as a function of each team's physical, (possibly deceptive) informational, and multi-robot capabilities.

Advisor
Kumar, Vijay
Pappas, George, J
Date of degree
2024
Date Range for Data Collection (Start Date)
Date Range for Data Collection (End Date)
Digital Object Identifier
Series name and number
Volume number
Issue number
Publisher
Publisher DOI
Journal Issue
Comments
Recommended citation