Attack Resilient State Estimation for Vehicular Systems

Source Organization:
University:University of Pennsylvania
Principal Investigator:Nicola Bezzo, Insup Lee
PI Contact Information:
Project Manager:
Funding Source(s) and Amounts Provided (by each agency or organization):
Total Dollars:
Agency ID/Contract/Grant Number:
Start and End Dates:
Project Status:Complete
Subject Categories:
Abstract:In recent years we have been witnessing an increase in autonomous vehicles: most of the cars we drive nowadays use multiple sensors to maintain constant speed (e.g., adaptive cruise control), avoid obstacles and collisions, park, move autonomously through traffic, and improve the overall driving comfort. In the presence of a cyber attack in which the received information from the sensors is compromised, safety is also compromised. Thus, our goal is to develop a resilient framework to guarantee vehicular safety in the presence of malicious cyber attacks. To consider cyber attacks that can compromise the overall safety of a vehicle, we are exploiting sensor fusion and redundancy in the sensor measurements. Specifically the problem under investigation is the following: Given a vehicle with N sensors measuring directly or indirectly a certain state, find the set of policies such that the vehicle can achieve a desired state while one or more sensor measurements are maliciously compromised by an adversarial attack. We focus primarily on control design schemes and address attacks on sensors for autonomous ground vehicles. We build upon ways to introduce redundancy within the control loop, as well as methods for attack detection and identification. We utilize security-aware attack-resilient estimators that identify an attack and allow the controller to pursue a mitigation strategy.
Describe Implementation of Research Outcomes (or why not implemented):
Impacts/Benefits of Implementation (actual, not anticipated):
Project URL: