An Evaluation System Based on User Big Data Management and Artificial Intelligence for Automatic Vehicles

Background

Safety issues have always been the focus of research on autonomous driving, so that the evaluation system has received increasing attention this year. J. Han et al. (2016) proposed a vision-based FCW and AEB system. In this paper, the author installed a laser scanner synchronously with the vision sensor, and the perception data of the laser scanner was used as the ground truth to evaluate the detection results of the system, thus being able to method can evaluate the accuracy of the detected distance. Lim et al. (2017) presented an integrated system of vehicle detection and distance estimation for a real-time AEB system based on stereo vision (Xie et al., 2017). Apart from that, the authors here use the KITTI dataset, which was co-founded by Karlsruhe Institute of Technology in Germany and Toyota American Institute of Technology, as a benchmark to evaluate the detection accuracy of the verification function and estimate the processing time of the algorithm on the Titan X platform and the NVIDIA TXI platform. An et al. (2019) put forward a framework for evaluating ADAS performance when considering four aspects of efficiency, safety, cost, and driver acceptance. Furthermore, macro-level factors (i.e., traffic volume and driving time) and micro-level indicators (i.e., reaction time and driver stress) were assessed. Due to limited time and experimental conditions, only ACC was tested as the main function of ADAS. However, the evaluation methods of the above systems consume a lot of time and human resources. In functional development, the evaluation scenarios are not sufficient, resulting in the risk of functional defects. Therefore, this study presents an evaluation system based on big data (Qi et al., 2015), aiming to reduce test costs and improve test reliability.