Safety and Security Assurance for Automotive Systems

Speaker: Pradeep Sharma Oruganti, mechanical engineering student

Abstract:
There has been increasing incorporation of connectivity and automation into systems that have been traditionally isolated and mechanical. From applications ranging from transportation to manufacturing, these large, interconnected systems are increasingly being tasked to perform several safety-critical functions. Considering the widespread implementation of these systems and increasing reliance on them, they have also become prime targets for malicious actors. In such scenarios, it is essential to develop and test systems to ensure safety and security right from the design stage of the product development cycle. This dissertation is an effort to develop tools and methodologies towards this goal. Given the scale of modern connected and autonomous systems, it is impossible to have a single solution for the entire system. To achieve the goals of this thesis, we split our study to target specific challenges seen at the network level, at the controller level, and in testing and validation. We draw from the fields of game theory, control theory and optimization to address issues in network design, safe control synthesis and safety validation of connected and autonomous vehicles. We begin by studying the problem of improving network level security under restricted budgets. We consider scenarios where the designer does not have access to the optimal location for investment and must intervene at other locations to strengthen the network. We propose algorithms to handle large inter-connected networks by reducing them to equivalent networks of lower dimensionality for faster computation. We further study various network design interventions to improve network security and showcase our proposed approaches by applying them on practical automotive and industrial systems. We follow this by focusing on low-level safety at the controller. We study the utility of approaches based on control barrier functions to ensure the safety of nonlinear dynamical systems in presence of unknown actuator inputs and bounded sensor noise. We expand the recently proposed high-order control barrier function to its measurement-robust variation and showcase the utility of our proposed approach by implementing it on a collision avoidance application. Next, we look briefly at the issue of safety validation of autonomous vehicles using scenario based testing approaches and provide a preliminary framework for large scale traffic simulation and hardware-in-loop based safety validation. Finally we end the dissertation by providing some conclusions and future directions.

Zoom Link (or alternative) - if available

Committee Members
Assistant Professor Qadeer Ahmed
Assistant Professor Parinaz Naghizadeh
The Ford Motor Company Chair in Electromechanical Systems Giorgio Rizzoni