Robust and Secure Design for Connected and Autonomous Vehicles

Modern vehicles are examples of complex cyber-physical systems with tens to hundreds of interconnected Electronic Control Units (ECUs) that manage various vehicular subsystems. The aggressive attempts of automakers in recent years to realize fully autonomous vehicles have led to the adoption of advanced machine learning based techniques for improved perception and control. Emerging vehicles are also becoming increasingly connected with various external systems (e.g., smart roadside units, other vehicles) to realize more robust vehicle autonomy. These paradigm shifts have resulted in significant overheads in resource constrained ECUs and increased the complexity of the overall automotive system, which has severe performance and safety implications in modern vehicles. The increased complexity introduces several computation and communication uncertainties in automotive subsystems that can cause real-time performance violations. Harsh operating conditions in vehicles further create a significant risk to the integrity of the data that is exchanged between ECUs which can lead to faulty vehicle control. The increased external connectivity also creates a large attack surface that is highly vulnerable to various kinds of sophisticated security attacks. In this talk, I will discuss the spectrum of reliability, security, and real-time performance challenges in contemporary automotive systems. I will then present a framework for efficient resource management in emerging automotive platforms that satisfies a diverse set of constraints related to reliability, security, real-time performance, and energy consumption. Several techniques inspired by advances in machine learning will be presented, with the goal of realizing robust and secure vehicle operation. Lastly, I will discuss open challenges and new directions of importance as we race towards an autonomous transportation future.