Speakers

Talk title: From provably safe to risk-aware robot planning and control  

Talk abstract: How can we ensure that autonomous robots work as expected and how can we even specify what it means to work as expected? Formal methods have shown to be useful in addressing these questions; temporal logics provide rich, rigorous, yet user-friendly specification formalism and formal synthesis offers a way to automatically generate a plan or a controller that provably satisfies the specification — under certain assumptions. However, these assumptions (e.g., on having perfect knowledge of the robot’s state) may be hard to achieve when moving into uncontrolled environments, and relaxing them may (e.g., by considering the worst-case scenarios) may leave us with the answer that provable, “perfect" safety is not achievable. In this talk, we focus on moving from provably safe to minimum-violation to risk-aware formal methods-based planning and control. We will discuss the use of temporal logics (LTL and STL) and their quantitative semantics in order to measure to what exten a robots meets a desired task; and show formal methods-based techniques that maximize a specification satisfaction.  

Bio: Jana Tumova is an associate professor at the School of Electrical Engineering and Computer Science at KTH Royal Institute of Technology. She received PhD in computer science from Masaryk University and was awarded ACCESS postdoctoral fellowship at KTH in 2013. She was also a visiting researcher at MIT, Boston University, and Singapore-MIT Alliance for Research and Technology. Her research interests include formal methods applied in decision making, motion planning, and control of autonomous systems. Among other projects, she is a recipient of a Swedish Research Council Starting Grant to explore compositional planning for multi-agent systems under temporal logic goals and a WASP Expeditions project focusing on design of correct-by-design and socially acceptable autonomous systems. She was selected to give an Early Career Spotlight talk at Robotics: Science and Systems 2021. 

Talk title: Role of Safety: from safety-critical control to safety-informed motion forecasting 

Talk abstract: The rapid advancement of machine learning and computation tools has brought promises of deploying fully autonomous robots beyond controlled factory floors. Ensuring their safe operation across various environments, particularly in uncertain, unseen, and unforgiving scenarios, is of paramount importance. In this talk, I will discuss the overarching concept of a safety filter: an automatic process that monitors the operation of an autonomous system at runtime and intervenes, when deemed necessary, by modifying its originally intended control to prevent a potential catastrophic failure. We will see that the guarantees provided by a safety filter depend significantly on its underpinning formal machinery. In the second part of the talk, we switch gears to look at the development of safety-informed human-centered interactive robot autonomy, specifically in the context of autonomous driving. I will show how the safety concepts discussed in the first half of the talk can be used to quantify the responsibility of road users in actual and counterfactual safety-critical events. I will demonstrate that the responsibility metrics can serve as an interpretable layer for an underlying trajectory predictor, which can integrate into the software stack for responsibility-aware autonomy.

Bio:  Kai-Chieh Hsu is a 5th-year Ph.D. candidate in electrical and computer engineering at Princeton University. He is fortunate to work with Prof. Jaime Fernández Fisac at Safe Robotics Lab. His research interests broadly lie in safe learning, safe multiagent planning, and prediction-planning integration. He obtained his B.S. in electrical engineering from National Taiwan University in 2019, where he worked on digital circuit design for healthcare and direction-of-arrival estimation.

Talk title: Traffic Law Abiding Autonomous Vehicles

Talk abstract:  Injecting traffic-law awareness in autonomous vehicle (AV) motion planners is crucial for ensuring their safe operation. A significant obstacle to achieving this is the complex nature of the traffic law which permits traffic-rule violation under certain exceptional situations. The sheer abundance of these exceptions makes a comprehensive classficiation and enumeration infeasible. In this talk, I will present an approach to express the traffic law in the form of a signal temporal logic (STL) rule hierarchy. I will delve into how this hierarchical representation can be effectively utilized for motion planning and elaborate on integrating these rule hierarchies with learning-based trajectory predictors.

Bio: Sushant Veer is a Research Scientist with the Autonomous Vehicle Research Group at NVIDIA Research. Broadly, his research interests lie in ensuring the safety of complex autonomous robotic systems. He is currently interested in improving the safety of autonomous vehicles by equipping them with the ability to detect and safely address edge cases that lie beyond the operational design domain. He was a Postdoctoral Research Associate in the Mechanical and Aerospace Engineering Department at Princeton University and received his Ph.D. in Mechanical Engineering from the University of Delaware in 2018 and a B.Tech. in Mechanical Engineering from the Indian Institute of Technology, Madras in 2013. In the past, He has worked on providing performance guarantees for learning-based motion planners, safe planning and control of dynamic-legged robots, and the development of assistive biomechanical devices.

Talk title: Safe Learning in Robotics: From Learning-based Control to Safe Reinforcement Learning

Talk abstract: Advancements in machine learning techniques have presented new opportunities for robotics, but applying these methods to real-world robotics applications remains challenging. Both the controls and the reinforcement learning communities have been addressing the challenge of safety in robot learning. Their efforts have been summarized in our recent review paper, “Safe Learning in Robotics: From Learning-based Control to Safe Reinforcement Learning.” In this talk, I will present the key findings of our review paper, highlight my team’s research in this context, and show experimental demonstrations of safe learning algorithms on flying robots, ground vehicles, and mobile manipulators. I will conclude by highlighting the many open questions in this field and hope to convince you to help solve these challenges. Finally, I will introduce the open-source simulation environment “safe-control-gym” developed by my team with the goal of facilitating the development of safe learning algorithms and accelerating progress in this field.

Bio: Angela Schoellig is an Alexander von Humboldt Professor for Robotics and Artificial Intelligence at the Technical University of Munich. She is also an Associate Professor at the University of Toronto Institute for Aerospace Studies and a Faculty Member of the Vector Institute in Toronto. Angela conducts research at the intersection of robotics, controls, and machine learning. Her goal is to enhance the performance, safety, and autonomy of robots by enabling them to learn from past experiments and from each other. In Canada, she has held a Canada Research Chair (Tier 2) in Machine Learning for Robotics and Control and a Canada CIFAR Chair in Artificial Intelligence, and has been a principal investigator of the NSERC Canadian Robotics Network. She is a recipient of the Robotics: Science and Systems Early Career Spotlight Award (2019), a Sloan Research Fellowship (2017), and an Ontario Early Researcher Award (2017). She is one of MIT Technology Review’s Innovators Under 35 (2017), a Canada Science Leadership Program Fellow (2014), and one of Robohub’s “25 women in robotics you need to know about (2013)”. Her team is the four-time winner of the North-American SAE AutoDrive Challenge (2018-21).

Her PhD at ETH Zurich (2013) was awarded the ETH Medal and the Dimitris N. Chorafas Foundation Award. She holds both an M.Sc. in Engineering Cybernetics from the University of Stuttgart (2008) and an M.Sc. in Engineering Science and Mechanics from the Georgia Institute of Technology (2007).