Colloquium Cognitive Systems

In the driver's mind: Cognitive modeling of human drivers in traffic interactions

PDr. A. Arkady Zgonnikov (online) (TU Delft, NL)

Abstract:
Human behavior models are critical for the development of automated vehicles (AVs): they are used for behavior prediction, interaction planning, virtual training and testing, and benchmarking of AVs. However, diverse models used for each of these applications often face similar challenges when it comes to accuracy, generalization, interpretability, and scalability. In this talk, I will argue that improving the ability of AVs to interact with humans requires fundamental scientific research into human cognitive processes during traffic interactions. In my lab we are developing a new generation of human driver behavior models to address this challenge. I will present an overview of our recent work covering several complementary approaches for modeling human drivers’ decision making and operational behavior in traffic interactions, and will illustrate the potential of these models for automated vehicles of the future.
About:
Arkady Zgonnikov is an Assistant Professor in the Department of Cognitive Robotics, Delft University of Technology. Trained in applied mathematics and computer science, he began his research career in computational cognitive science, focusing on human decision making and motor behavior. His current work centers around understanding and modeling human behavior in traffic interactions and applying these insights to the development and evaluation of automated vehicles. He is also actively advancing the concept of meaningful human control over intelligent systems, both as a theoretical framework and as a practical principle for AI design and assessment (with a particular focus on automated vehicles). He serves as Director for Partnerships and Collaborations at the Centre for Meaningful Human Control, and as Co-Director of the HERALD Lab (Human-Aware Robust AI for Automated Driving).