Learning and Control for Safety-Critical Systems

The last decade has witnessed a dramatic increase in the presence of autonomous systems in many domains, ranging from transportation networks to defense systems and healthcare services. Many of these systems are built using costly hardware and operate in uncertain and dynamically changing environments while interacting with humans and other agents. As potential failures can have catastrophic economic and societal consequences, the reliable deployment of these autonomous systems critically depends on the verifiability of the systems’ safety and security.

Assuring certifiably safe autonomy requires a number of challenges to be addressed. The broad challenges that will be discussed in this workshop include, but are not limited to,

• safe and efficient decision-making under limited information,

• on-the-fly system identification and control with safety guarantees,

• planning and learning in adversarial environments, and

• distributed and resilient learning in multi-agent systems.