To ensure the reliability and safety of autonomous driving systems, collecting high-quality training data is crucial. Given the challenges and costs of real-world road data collection, simulation technology has become a key means of training and testing autonomous driving algorithms. Compared to traditional simulators, scene generation based on rendering technology is more realistic when handling extreme and complex scenarios. In recent years, NeRF (Neural Radiance Fields) technology has provided a new solution for the reconstruction of autonomous driving simulation scenes. NeRF technology uses neural networks to represent and render 3D scenes, allowing high-quality 3D representations to be recovered from 2D images captured from multiple viewpoints, creating highly realistic scene effects. Applying NeRF technology to autonomous driving simulation scenes not only enables the high-quality reconstruction of complex driving environments and accurate reproduction of various situations under harsh conditions but also reduces the difficulty and cost of data collection, while effectively validating the performance of autonomous driving algorithms. This paper will explore the prospects of NeRF technology in autonomous driving simulation scenes, with a focus on outlining the basic principles of NeRF technology and discussing the current state of NeRF-based reconstruction for autonomous driving simulation scenes.