Reilient Multi-Vehicle Networks
Abstract
The focus of this proposal is on the modeling, analysis, detection and alleviation of cyberattacks in swarms of autonomous unmanned aerial vehicles (UAVs). The class of cyberattacks considered are those wherein an attacker hacks into the guidance & control program of a carefully chosen subset of vehicles within the swarm, and converts them into vehicles with malicious intent. These malicious vehicles then perform a series of subtle changes in the way they interact with the other vehicles in the swarm, with the intention to degrade the mission effectiveness of the swarm. Similar attacks may also be performed on autonomous self-driving cars of the future. By hacking into the driving software of a carefully chosen subset of these vehicles, and performing subtle changes in their driving behavior, an attacker can create shock waves on highways, and such shock waves can lead to pile-up crashes. The proposed research seeks to make vehicle networks resilient to such attacks. Towards this end, the proposal uses a Partial Differential Equation (PDE) framework for modeling such scenarios. This PDE framework comprises of two species of vehicles: malicious vehicles and normal vehicles, and the malicious vehicles may be arbitrarily interspersed among the normal vehicles. The advantages of such a PDE framework are that it is scalable to an arbitrarily large number of vehicles in the swarm, and it also enables a spatio-temporal characterization of the swarm behavior. This PDE framework is employed to meet the following objectives: (a) Perform a rigorous mathematical modeling of such attacks, (b) Use these models as a foundation to develop novel estimation schemes that can detect such attacks as they are in progress (c) Use decision-making methods to determine suitable countermeasures to foil these attacks, and analyze these countermeasures in a game-theoretic framework (d) Develop a comprehensive software-in-the-loop testbed to validate the performance of the above algorithms through extensive simulations in nominal as well as off-nominal conditions. The outcomes of this research will benefit homeland security by ensuring safe and resilient autonomous operation of swarms in complex environments, as well as contribute to the development of smart cities, by making self-driving cars resilient to attacks. The students working on this project will be trained to perform research at the novel intersection of PDE modeling, cybersecurity, multi-agent systems, computer vision and cooperative control. Important aspects from this research will be integrated into several courses taught by the PIs, thereby training a large cohort of students to develop critical analytical and computational skills. The PIs and their students will conduct regular school visits and webinars to demonstrate the positive impact of STEM on addressing problems impacting society and homeland security. Development of the comprehensive software-in-the-loop testbed that integrates PDE models of swarms, multi-vehicle dynamics, intervehicle communication and computer vision algorithms will significantly enhance the research capabilities of the PIs institution.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 24, 2023
- Source ID
- W911NF2310174
Entities
People
- Animesh Chakravarthy
Organizations
- Army Contracting Command
- Office of the Secretary of Defense
- University of Texas at Arlington