Speaker: Dr. Tomer Weiss, New Jersey Institute of Technology
Collective behavior is all around us, from crowd rallies, ant and robot swarms, to avatars in virtual reality. Understanding such behavior is critical when preparing for crowd panic behavior in emergency situations, increasing the fidelity of virtual worlds, conceptualizing urban layouts, and improving human-robot cooperation. In this talk, I will present our approach for learning, modeling and simulating multi-agent dynamics.
First, I will demonstrate a position-based method for multi-agent simulation, which is an alternative to traditional velocity-obstacle approaches in robotics. Our method simulates large masses in interactive rates for hundreds of thousands of agents, which was previously unachievable. This work received the best paper award in the ACM SIGGRAPH conference on Motion in Games 2017. Second, I will discuss our recent results on Deep Reinforcement Learning for multi-agent navigation, which allows fine-grain rewards-based strategies for controlling agent locomotion behaviors. We will illustrate how our techniques can be applied to the simulation of human-like crowds, with applications to computer animation, gaming, pedestrian dynamics, visual surveillance and robotics.
Tomer Weiss is an assistant professor with the Department of Informatics, at the New Jersey Institute of Technology. Before that, he was a research scientist with Wayfair, Boston. In 2018, he defended his PhD advised by Prof. Terzopoulos at the University of California, Los Angeles. He received the Best Paper Award from the ACM SIGGRAPH conference on Motion in Games, for his work on virtual crowd simulation. He was a finalist presenter in both ACM SIGGRAPH Thesis Fast Forward, and the ACM SIGGRAPH Asia Doctoral Symposium in 2018. He received his MS in computer science from UCLA in 2016, and his BSc degree in computer science from Tel Aviv University in 2013. His research interests include collective dynamics, scene understanding, and interactive visual computing.