urban_road_filter:
Real-Time LIDAR-Based Urban Road and Sidewalk Detection for Robots and Vehicles

Széhenyi István University
2022-2026

*Indicates Equal Contribution

An example of real-time algorithm output

Abstract

urban_road_filter is an open-source package built on a set of algorithms presented across multiple papers. Road and sidewalk detection in urban scenarios is a challenging task due to road surface imperfections and high sensor data bandwidth. Traditional free-space and ground filter algorithms are not sensitive enough to detect small height differences. Camera-based or sensor-fusion approaches are widely used to classify drivable road from sidewalk or pavement, but a LIDAR sensor alone already contains all the information needed for feature extraction. This work therefore focuses on LIDAR-based feature extraction, presenting a real-time (20 Hz+) solution for road and sidewalk detection that can also be used for local path planning. Sidewalk edge detection is performed by combining three algorithms running in parallel. Results are validated against the de facto standard benchmark dataset, KITTI, alongside our own measurements. Data and source code to reproduce the results are shared publicly on our GitHub repository.

BibTeX

If you use this work, please cite one of the following:
@Article{roadfilt2022horv,
    title = {Real-Time LIDAR-Based Urban Road and Sidewalk Detection for Autonomous Vehicles},
    author = {Horváth, Ernő and Pozna, Claudiu and Unger, Miklós},
    journal = {Sensors},
    volume = {22},
    year = {2022},
    number = {1},
    url = {https://www.mdpi.com/1424-8220/22/1/194},
    issn = {1424-8220},
    doi = {10.3390/s22010194}
}