SYNOPSIS
Self-driving cars are expected to go beyond controlled environments, and soon, they will be deployed in public streets. Concerning with safe deployment of self-driving cars at a large scale, a major challenge is handling unexpected scenarios which primarily root in different sources of uncertainty coexist in practical realworld situations. Thus, there is an urgent need for developing motion planning techniques that can assure the safety of self-driving vehicles, particularly in the presence of uncertainty. Mathematically, this problem can be framed as a partially observable Markov decision process (POMDP). The optimal behavior in a POMDP domain is expected to strike a balance between exploring the partially observable world and acting in a goal-directed manner. However, POMDPs are notoriously difficult to solve when the number of states is even moderately large because the solution algorithm efficiency decays exponentially with the size of the state space. This talk will cover several safe and scalable policy synthesis techniques for large-scale POMDPs. The central theme of this talk is to search for “reliable and efficient planning algorithms” to bridge the gap between theoretical developments and real-world applications.
ABOUT THE SPEAKER
Abdullah Al Redwan Newaz is a Postdoctoral Research Associate at Autonomous Cooperative Control of Emergent Systems of Systems (ACCESS) Laboratory at North Carolina A&T State University. Before that, he was a Postdoctoral Researcher at Rice University, Houston, TX, USA, and Nagoya University, Nagoya, Japan, in 2018-2020, 2017-2018, respectively. He earned Ph.D. and M.S. degrees in Information Science from Japan Advanced Institute of Science and Technology, Ishikawa, Japan, in 2017 and 2014, respectively. He received a B.Sc. degree in Mechanical Engineering from Rajshahi University of Engineering and Technology, Rajshahi, Bangladesh, in 2011. His research interests include robotics, artificial intelligence, autonomous systems, reinforcement learning, deep learning, computer vision, behavior planning, optimal control, policy or controller synthesis, formal methods, and model checking. He is an IEEE professional member and serves as an active reviewer for robotics, control, and transportation conferences and journals.