Adaptive Design of Experiments for Accelerated Safety Evaluation of Automated Vehicles

Automated Vehicles (AVs) need to be thoroughly evaluated in order to ensure their driving capabilities. However, comprehensive evaluations are intractable due to both time and monetary costs. To address this problem, we propose an Adaptive Design of Experiments (ADOE) method to evaluate the safety of AVs. For this method, a surrogate model is established iteratively and is used to approximate the results of AV testing. During each iteration, the next concrete scenario to be tested is selected with the guidance of the surrogate model. Since the choice of the surrogate model has a profound impact on the performance of this proposed method, 6 surrogate models were compared with two logical scenarios at different scales. Results show that the Extreme Gradient Boosting (XGB) model stands out amongst popular surrogate models. Compared to brute-force enumeration, the proposed ADOE method based on XGB captured 94% of concrete collision cases while only using 4.5% of computational resources. The proposed method has great potential in accelerating the evaluation process of AVs.