Autonomous Driving with Deep Learning: A Survey of State-of-Art Technologies

Yu Huang,Yue Chen

2020-07-04

Abstract:Since DARPA Grand Challenges (rural) in 2004/05 and Urban Challenges in 2007, autonomous driving has been the most active field of AI applications. Almost at the same time, deep learning has made breakthrough by several pioneers, three of them (also called fathers of deep learning), Hinton, Bengio and LeCun, won ACM Turin Award in 2019. This is a survey of autonomous driving technologies with deep learning methods. We investigate the major fields of self-driving systems, such as perception, mapping and localization, prediction, planning and control, simulation, V2X and safety etc. Due to the limited space, we focus the analysis on several key areas, i.e. 2D and 3D object detection in perception, depth estimation from cameras, multiple sensor fusion on the data, feature and task level respectively, behavior modelling and prediction of vehicle driving and pedestrian trajectories.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to effectively integrate deep - learning methods in autonomous driving technology to improve the performance of autonomous driving systems. Specifically, the paper focuses on the following aspects:

1. **Perception**: How to use deep - learning techniques for environmental perception, including 2D/3D object detection, depth estimation, multi - sensor fusion, etc. These techniques can help vehicles more accurately identify objects in the surrounding environment, such as pedestrians, other vehicles, traffic signs, etc.

2. **Mapping and Localization**: How to construct high - precision maps through deep - learning methods and achieve accurate vehicle localization. This involves using data from multiple sensors such as cameras, radars, LiDARs, as well as SLAM (Simultaneous Localization and Mapping) technology.

3. **Prediction**: How to predict the motion trajectories of other traffic participants, including behavior modeling of vehicles and pedestrians. This helps vehicles to react in advance and avoid potential dangers.

4. **Planning and Control**: How to use deep - learning techniques for path planning and vehicle control. This includes from global path planning to local behavior decision - making, and then to the generation of specific vehicle control instructions.

5. **Simulation**: How to construct a realistic simulation environment in order to fully test and verify the autonomous driving system before actual road tests.

6. **V2X Communication**: How to use vehicle - to - vehicle (V2V) and vehicle - to - infrastructure (V2I) communication technologies to improve the safety and efficiency of autonomous driving.

7. **Safety**: How to ensure the functional safety of the autonomous driving system, especially the robustness in the face of uncertainty and complex environments.

Through in - depth investigation and analysis of these fields, the paper aims to provide comprehensive references and guidance for the development of autonomous driving technology.