Abstract: How will self-driving cars change urban mobility? This talk describes both theoretical and scientific contributions in the field of reinforcement learning, combinatorial optimization, and convex optimization, presented in the context of enabling mixed-autonomy mobility, the gradual and complex integration of automated vehicles into the existing traffic system.
The talk first presents new contributions in mixed-autonomy traffic, which explore and quantify the potential impact of a small fraction of automated vehicles on low-level traffic flow dynamics, using novel techniques in model-free deep reinforcement learning. Second, the talk presents reinforcement learning techniques for improved variance reduction developed for the control of complex large-scale systems such as mixed-autonomy traffic presented in the first part. To anchor this work in a broader mobility planning perspective, recent theoretical contributions in convex and combinatorial optimization, which contribute to challenges in higher-level individual and aggregate decision making in mixed-autonomy mobility will then be presented in the third part of the talk. More generally, the continued and accelerating introduction of automation into society will generate widespread yet poorly understood externalities. Therefore, this talk concludes by proposing a multi-scale design paradigm for developing sustainable and reliable intelligent infrastructure, which extends the current methodology for building AI systems, and a taxonomy of critical challenges both in machine learning and modeling, needed to address growing problems created by rapid urbanization.
Bio: Cathy Wu is a PhD candidate in machine learning in Electrical Engineering and Computer Science (EECS) at UC Berkeley, Berkeley Artificial Intelligence Research (BAIR), Berkeley DeepDrive (BDD), California PATH, and the Berkeley RISELab. She received a Master of Engineering degree in EECS (2013) and a Bachelor of Science degree in EECS (2012) from the Massachusetts Institute of Technology (MIT).
Cathy Wu is the recipient of several fellowships including NSF Graduate Research Fellowship, the Berkeley Chancellor's Fellowship for Graduate Study, the NDSEG Fellowship, and the Dwight David Eisenhower Graduate Fellowship. Her work was acknowledged by several awards, including the 2016 IEEE ITSC Best Paper Award and the ITS Outstanding Graduate Student Award. Her leadership, in particular as the Lead of the Learning Traffic Research Team at UC Berkeley and the Chair of the Interdisciplinary Research Initiative within the ACM Future of Computing Academy, was recognized by numerous awards and invitations, such as multiple NSF early-career investigator workshops on cyber-physical systems and the 2017 IEEE Leaders Summit. Her areas of interest include machine learning, optimization, and control systems, particularly problems concerning network dynamical systems, multi-agent learning systems, and transportation. Her research goal is to advance computational methods to enable intelligent infrastructure. http://www.wucathy.com/blog/