Development of an Integrated Augmented Reality Testing Environment and Implementation at the American Center for Mobility

Liu, Henry · 2023 · ROSA P / University of Michigan. Center for Connected and Automated Transportation

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the critical challenge of efficiently and accurately testing autonomous vehicles (AVs) by developing an integrated testing environment that combines a high-fidelity Naturalistic Driving Environment (NDE) with Augmented Reality (AR) technology. Traditional real-world testing is prohibitively expensive and inefficient due to the rarity of safety-critical events, while existing simulation methods often suffer from a "simulation-to-reality gap" because they fail to accurately model the stochastic, distributionally consistent behaviors of human-driven background traffic. The authors propose a data-driven, optimization-based framework to generate virtual traffic that statistically mirrors real-world driving distributions, thereby enabling reliable AV safety evaluation. The methodology involves a two-step NDE modeling process validated using large-scale naturalistic driving data from the University of Michigan’s IVBSS and SPMD datasets. First, empirical behavior models for longitudinal (free-driving, car-following) and lateral (lane-changing) actions are constructed directly from observed driving data. Second, to correct for error accumulation inherent in empirical models, the authors apply an optimization framework based on Markov chain stationary distributions. This robust modeling step adjusts the empirical models to ensure the simulated environment’s long-term velocity and range distributions match real-world ground truth. The resulting NDE is integrated with an AR system that synchronizes simulation data with the physical world, rendering virtual background vehicles visible to the AV under test. This integrated solution was implemented at the American Center for Mobility (ACM) in Michigan. The results demonstrate that the proposed method significantly outperforms existing baselines, such as the SUMO simulator and Intelligent Driving Model (IDM), in reproducing real-world velocity and range distributions. Quantitative evaluation using Hellinger distance confirms superior distributional consistency. In AV safety testing, the proposed NDE generated a measurable accident rate of $5.5 \times 10^{-5}$ per simulation, producing 276 diverse safety-critical events, whereas the SUMO baseline yielded zero accidents due to its conservative, accident-free design. The system successfully facilitated real-time testing at ACM, allowing the AV to interact with virtual traffic that appeared physically present. The significance of this work lies in providing a scalable, cost-effective solution for AV validation that bridges the gap between simulation and reality. By ensuring distributional consistency in background traffic behavior, the framework enables statistically accurate safety performance estimation without requiring billions of miles of real-world driving. The implementation at ACM demonstrates the practical viability of combining data-driven NDE modeling with AR, offering a robust platform for evaluating AV safety with high fidelity and operational efficiency.

Key finding

The proposed data-driven and optimization-based NDE modeling framework produces simulation environments with significantly higher distributional consistency to real-world driving than existing models, enabling more accurate and efficient autonomous vehicle safety testing.

Methodology

mixed_methods

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• situational awareness

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Methodological Resource: tool software, validation psychometrics
• Theoretical Contribution: computational model