Driving intelligence tests are critical to the development and deployment of autonomous vehicles. The prevailing approach tests autonomous vehicles in life-like simulations of the naturalistic driving environment. However, due to the high dimensionality of the environment and the rareness of safety-critical events, hundreds of millions of miles would be required to demonstrate the safety performance of autonomous vehicles, which is severely inefficient. We discover that sparse but adversarial adjustments to the naturalistic driving environment, resulting in the naturalistic and adversarial driving environment, can significantly reduce the required test miles without loss of evaluation unbiasedness. By training the background vehicles to learn when to execute what adversarial maneuver, the proposed environment becomes an intelligent environment for driving intelligence testing. We demonstrate the effectiveness of the proposed environment in a highway-driving simulation. Comparing with the naturalistic driving environment, the proposed environment can accelerate the evaluation process by multiple orders of magnitude.

No abstract

No abstract

Microscopic traffic simulation provides a controllable, repeatable, and efficient testing environment for autonomous vehicles (AVs). To evaluate AVs' safety performance unbiasedly, ideally, the probability distributions of the joint state space of all vehicles in the simulated naturalistic driving environment (NDE) needs to be consistent with those from the real-world driving environment. However, although human driving behaviors have been extensively investigated in the transportation engineering field, most existing models were developed for traffic flow analysis without consideration of distributional consistency of driving behaviors, which may cause significant evaluation biasedness for AV testing. To fill this research gap, a distributionally consistent NDE modeling framework is proposed. Using large-scale naturalistic driving data, empirical distributions are obtained to construct the stochastic human driving behavior models under different conditions, which serve as the basic behavior models. To reduce the model errors caused by the limited data quantity and mitigate the error accumulation problem during the simulation, an optimization framework is designed to further enhance the basic models. Specifically, the vehicle state evolution is modeled as a Markov chain and its stationary distribution is twisted to match the distribution from the real-world driving environment. In the case study of highway driving environment using realworld naturalistic driving data, the distributional accuracy of the generated NDE is validated. The generated NDE is further utilized to test the safety performance of an AV model to validate its effectiveness.

Signal phase and timing (SPaT) information is critical for many in-vehicle applications. However, it is challenging and time-consuming to acquire city-wide SPaT information from local traffic management agencies directly. A significant limitation of existing SPaT information estimation methods in the literature is that they can only be applied to a specified time-of-day (TOD) period. In the real-world, however, different TOD timing plans are used to accommodate fluctuations in traffic demands. In this paper, we propose a novel method for traffic light parameter estimation based on floating car data, which features recognizing TOD breakpoints and can thus be applied to intersections with multi-TOD timing plans. Also, good estimation of TOD breakpoints leads to more data availability for estimation of other parameters. The proposed method is tested with real-world data collected from the DiDi on-line hailing platform in China. The filed test results show promising accuracy. The absolute error of green duration is within 3 s in daytime and the estimation error of TOD breakpoints is within 15 min.