Overview

Title

Collaborative Research: OAC Core: Stochastic Simulation Platform for Assessing Safety Performance of Autonomous Vehicles in Winter Seasons (2106991 and 2106965)

PI: Xianfeng Yang, University of Maryland, xtyang@umd.edu

co-PI: Xiaobai Liu, San Diego State University, xiaobai.liu@sdsu.edu

Besides the accuracy of object detection, the safety performance of an Autonomous Vehicle (AV) alsohighly depends on the reliability of its Automated Driving System (ADS) algorithms. Despite the success of AVs in reducing crash risks have been proved by many field tests, their safety performance, under adverse driving conditions in winter seasons, still lacks comprehensive evaluations. Hence, to avoid costly mistakes before the widespread implementations, there is an urgent need to build a reliable cyberinfrastructure, a stochastic simulation platform more specifically, to pre-evaluate the capability of ADS algorithms in dealing with icy/snowy driving conditions. To overcome those challenges, our research plan includes the following research thrusts: 1) stochastic simulation with vehicle behavioral modeling; 2) AV Safety assessment with vehicle dynamic modeling; 3) platform development with incremental online learning; and 4) model validation and transition-to-practice plan development. The proposed research is creative and original as it is the first study, based on the principle of physics regularized Gaussian process (PRGP), to model vehicle behaviors on the icy/snowy pavement. Furthermore, the proposed cyberinfrastructure will provide an open-source and public simulation platform that grants access to the entire AV community. The embedded incremental online learning function can facilitate the gathering of much more diverse training data from users and ensure the system’s sustainability beyond the life of this research grant.