< Back to Session View

Using Human-in-the-Loop Simulations to Study Impacts of Automated Vehicle in Urban and Highway Driving

Date and Time: Tuesday, July 11, 2023: 5:30 PM - 7:00 PM

Xingyu Zhou 

Graduate Research Assistant, University of Texas at Austin

Presentation Description

The University of Texas at Austin Center for Transportation Research, in collaboration with the Texas Department of Transportation (TxDOT), conducted research to address concerns about the safety of automated vehicles (AVs) on public roadways. The team developed an operational design domain (ODD) architecture that considers the various factors that affect AV operations, such as traffic characteristics, roadway geometry, infrastructure quality, environmental and weather conditions, geographic constraints, and objects and events. The ODD architecture is intended to help improve the safety of AVs by ensuring that they operate within a controlled and repeatable environment. To test the performance of AVs within the ODD, the research team used a high-fidelity moving-base driving automation simulator, which provided a immersive yet safe way to compare the driving behavior, safety, and comfort of AVs and human drivers. The Level-3 AVs were equipped with Lane Departure Warning, Lane Keeping Assist, Lane-Centering Automatic Steering Control, and Adaptive Cruise Control. The team developed four driving scenarios that prioritized challenging roadway environments such as work zones, on- and off-ramps, and adverse weather conditions. TxDOT ultimately selected two ODDs for simulation: highway and urban. In the highway scenario, the ego vehicle enters an on-ramp and merges into the travel main lanes, and the simulated vehicle merges into the ego vehicle’s travel lane. In the urban scenario, the ego vehicle merges due to a work zone, and the simulated vehicle merges into the ego vehicle’s travel lane due to the work zone. The research team used RoadRunner-Unity software to build the ODD environments, which included several simulated vehicles and seven full-dynamic vehicles that react in real-time to the ego vehicle’s behavior. The research team recruited 20 participants to test the scenarios, with each participant assigned one scenario to test in both automated and manual modes. Following the test, participants completed a survey that provided input regarding their familiarity with AVs and feedback on their experience with the simulator. The results of the study help to improve our understanding of AV operations and provide insights into how to address the concerns of consumers and policymakers regarding AV safety performance.

Speaker Biography

Xingyu Zhou received his B.S. degree (with Highest Distinction) in mechanical engineering from Purdue University, West Lafayette, IN, USA, in 2016, and the M.S. degree in mechanical engineering from the University of Michigan, Ann Arbor, MI, USA, in 2018. Since 2020, he has been working towards a Ph.D. degree in the Walker Department of Mechanical Engineering, The University of Texas at Austin, Austin, TX, USA.

His general research interests lie in the intersection of advanced control engineering, machine learning, optimization, and mechatronics. His current research focus is to tackle a wide spectrum of automotive challenges (human-automation synergy, human-centric automated driving technology, transportation electrification, intelligent transportation systems, etc.) leveraging the interplay between advanced control theories and machine learning methods.

Prior to his doctoral study, he worked in the industry for two years (2019 - 2020) and was a full-time control engineer. During this time, he specialized in advanced vehicle motion control system architecture and algorithm development, simulation platform design (SIL and HIL), embedded software development, and hardware/software bring-up and integration.

Presentation File