Contents I. Introduction
As one of the hottest research areas, automated driving is mainly contained computer vision (CV) and artificial intelligence (AI) technologies. In recent years, with the advance of fifth generation (5G) and sensor technology, detection, localization, cloud computing, etc., have also been rapidly developed. Although some autonomous vehicles and systems are already available on the market, some recent crashes highlighted that further research and more testing are necessary. In the whole automated driving system, recognition is a significant part therefore the choice of external sensors becomes important. For a full vision system, cameras can work at high resolution and have good recognition performance. While the power-up and cost decrease of computing makes that can run complex neural network models in real-time, the influence of weather conditions also makes it hard to achieve commercial level robustness. In addition, that difficult to fully fit using a finite mathematical model (neural network model) under the complexity and variability of real roads situation. Then, large computational will greater burden on the electrical system, especially for EV, and reduces the endurance capacity.
