Autonomous Driving in Unstructured Environments: How Far Have We Come?

Chen Min,Shubin Si,Xu Wang,Hanzhang Xue,Weizhong Jiang,Yang Liu,Juan Wang,Qingtian Zhu,Qi Zhu,Lun Luo,Fanjie Kong,Jinyu Miao,Xudong Cai,Shuai An,Wei Li,Jilin Mei,Tong Sun,Heng Zhai,Qifeng Liu,Fangzhou Zhao,Liang Chen,Shuai Wang,Erke Shang,Linzhi Shang,Kunlong Zhao,Fuyang Li,Hao Fu,Lei Jin,Jian Zhao,Fangyuan Mao,Zhipeng Xiao,Chengyang Li,Bin Dai,Dawei Zhao,Liang Xiao,Yiming Nie,Yu Hu,Xuelong Li

2024-11-01

Abstract:Research on autonomous driving in unstructured outdoor environments is less advanced than in structured urban settings due to challenges like environmental diversities and scene complexity. These environments-such as rural areas and rugged terrains-pose unique obstacles that are not common in structured urban areas. Despite these difficulties, autonomous driving in unstructured outdoor environments is crucial for applications in agriculture, mining, and military operations. Our survey reviews over 250 papers for autonomous driving in unstructured outdoor environments, covering offline mapping, pose estimation, environmental perception, path planning, end-to-end autonomous driving, datasets, and relevant challenges. We also discuss emerging trends and future research directions. This review aims to consolidate knowledge and encourage further research for autonomous driving in unstructured environments. To support ongoing work, we maintain an active repository with up-to-date literature and open-source projects at: <a class="link-external link-https" href="https://github.com/chaytonmin/Survey-Autonomous-Driving-in-Unstructured-Environments" rel="external noopener nofollow">this https URL</a>.

Robotics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenges faced in implementing autonomous driving technology in unstructured environments. Specifically, the paper points out that, compared with structured environments such as cities, unstructured environments such as rural areas and rugged terrains have higher complexity and diversity. These environments lack clear road signs, traffic signals and lane markings, posing more severe challenges to autonomous driving technology. These problems are mainly reflected in the following aspects:

1. **Diversity of environmental types**: There are significant differences in the terrain, landforms and features of unstructured environments such as grasslands, deserts, forests and mountainous areas, and changes in seasons and weather will further affect the geometric structure and appearance of the environment.

2. **A wide variety of elements**: Natural objects in unstructured environments are not only numerous in type but also diverse in form, and some elements are even difficult to distinguish manually.

3. **Ambiguity of semantic categories**: In unstructured environments, many environmental features are often intertwined, with complex spatial relationships and semantic connections, and blurred boundaries, making it difficult to accurately define the semantic category to which a specific element belongs.

4. **Disordered scene structure**: Unstructured environments lack clear traffic features such as lane markings and paved roads. The passable areas are irregular in shape, with unclear directions and variable structures, and usually there are no obviously recognizable roads.

5. **Complex road conditions**: Road surface materials in unstructured environments are diverse, including soil, gravel, grass and mud. Roads are often rough, with obvious vertical undulations or lateral slopes, and the geometric features of roads such as curvature and gradient are also more complex and variable.

6. **Fragile satellite signals**: In complex unstructured environments such as canyons and forests, satellite signals are easily blocked or attenuated, and the signals may be reflected or scattered during transmission, resulting in a significant decrease in positioning accuracy. Electromagnetic interference, which is common in military application scenarios, also seriously disrupts satellite signals.

By reviewing more than 250 related literatures, covering aspects such as offline map construction, pose estimation, environmental perception, path planning, end - to - end autonomous driving, data sets and related challenges, the paper aims to integrate knowledge and encourage further research to overcome the above challenges and promote the development of autonomous driving technology in unstructured environments.