The emergence of autonomous vehicles driving along human-driven traditional vehicles creates a mixed traffic environment on the roads. It is necessary to understand the intricate dynamics between the AVs and TVs while operating in terms of car-following. However, most of the AV car-following models are calibrated and validated based on homogenous traffic ignoring the mixed traffic dynamics. In this seminar, Dr. Das and Dr. Samandar delve into the complexities of car-following dynamics in mixed-traffic environments, where autonomous vehicles interact with traditional vehicles. The speakers introduce a cutting-edge car-following model, leveraging a long short-term memory (LSTM) based deep neural network specifically tailored for AVs navigating mixed traffic conditions. What sets this model apart is the incorporation of L2 regularization, a modification designed to enhance generalizability. This enhancement proves pivotal in accurately predicting high accelerations and decelerations, contributing to the model's heightened accuracy.
Dr. Tanmay Das, Ph.D. in Civil Engineering (Transportation Systems) from North Carolina State University, specializes in applying AI to traffic flow theory. His research centers on car-following dynamics and safety evaluations of autonomous vehicles (AVs) navigating mixed traffic scenarios with connected and autonomous vehicles (CAVs) and traditional vehicles (TVs). Currently serving as a Highway Safety Analyst at VHB in Raleigh, NC, Dr. Das also holds a role as a Research Assistant Scholar at the Institute for Transportation Research and Education, North Carolina State University.
Dr. Shoaib Samandar, a research scholar at NC State University, specializes in Intelligent Transportation Systems, focusing on the integration of artificial intelligence in transportation systems and the development of connected-autonomous vehicles. His research contributions in these domains are documented in over a dozen peer-reviewed publications.