Towards Trustworthy Swarming of Autonomous Vehicles
Abstract:
Though research in Cyber Physical Systems (CPS) continues to push forward the capabilities of autonomous robotic systems in many military and civilian applications, there exist tasks that require the use of more than a single robot. Such tasks can be addressed with swarms of autonomous robots. There have been several studies on the topic of robot swarms in the last few years, however these studies are mostly centered around motion planning and coordination design. Swarm technology is improving rapidly and will soon find its way into society. With this possibility in the near future it becomes critical to consider security of these systems to guarantee safety. The work presented in this thesis develops around these premises and proposes a framework to detect and overcome cyber-attacks in robotic swarms. The proposed framework utilizes virtual physics and graph theory rules to create a decentralized mesh of robotic vehicles. By leveraging inter-swarm interactions between neighbors in the swarms we then propose a trust-based detection algorithm to identify spoofing and hijacking attacks. Finally, we propose a checkpointing protocol to recover the state of attacked victim nodes.
Committee:
- Jack Davidson (Chair)
- Nicola Bezzo (Advisor)
- Yonghwi Kwon