Quantification of Actual Road User Behavior on the Basis of Given Traffic Rules

Code and supplementary materials

Repository of the paper Quantification of Actual Road User Behavior on the Basis of Given Traffic Rules at IV 2022. The first authors are strong supporters of the Public Money, Public Code idea and happy to contribute to it with this repository.

It provides two scripts to extract and learn the rule conformity of two traffic rules out of the Waymo Open Dataset Motion:

• Distance keeping
• Speed limitation

Each rule conformity is a percentage, that indicates how strongly a rule is followed among every driver in the dataset.

Results of both rules:

📋 Distance Keeping average rule conformity: 0.867

Statistics

Settings

• latency = 3
• step_size = 10
• size = None
• save_detail = True
• min_speed = 5.0
• anlge_range = 20.0

📋 Speed Limitation average rule conformity: 0.967

Statistics

Settings

• dist = 10.0
• step_size = 10
• size = None
• save_detail = True
• min_speed = 0.8

Install Repository

Install inside of a virtual environment (recommended)

pip install --upgrade pip
git clone https://github.com/fzi-forschungszentrum-informatik/quantification_road_user_behavior.git
pip install -r requirements.txt

Distance Keeping

For extracting the distance rule conformity run:

cd scripts
python DistanceRule.py --path path/to/dataset

which provides the following flags:

Flag	Description	Default
--path	Path to dataset	data/dataset/
--latency	Time of latency for the three-second rule	3
--step_size	Extract every [number]th state	10
--min_speed	Ignore drivers below this threshold	5.0
--angle_range	Prevent rule violation of a driver that is not facing an other vehicle among (+/-) [angle_range] (degree). Prevents violation in curves.	10
--size	Extract only a certain amount of files	None (=all files)
--save_freq	Update and save overall file	100
--save_detail	If True, save every state and the average file of all scenarios	True
--sample	If True, plot a random scenario with visualizations	False

Visualization of a random file among the dataset:

Speed Limitation

For extracting the Speed Limit rule conformity run:

cd scripts
python VelocityRule.py --path path/to/dataset

which provides the following flags:

Flag	Description	Default
--path	Path to dataset	data/dataset/
--dist	Maximum distance between driver and closest lane_center	10.0
--step_size	Extract every [number]th state	10
--min_speed	Percentage of the assigned speed limit. Discard drivers below this threshold	0.8
--size	Extract only a certain amount of files	None (=all files)
--save_freq	Update and save overall file	100
--save_detail	If True, save every state and the average file of all scenarios	True
--sample	If True, plot a random scenario with visualizations	False

Visualization of a random file among the dataset:

Citation

If you find this code useful for your research, please cite our paper:

@InProceedings{Bogdoll_Quantification_2022_IV,
  author    = {Bogdoll, Daniel and Nekolla, Moritz and Joseph, Tim and Z\"{o}llner, J. Marius},
  title     = {{Quantification of Actual Road User Behavior on the Basis of Given Traffic Rules}},
  booktitle = {IEEE Intelligent Vehicles Symposium (IV)},
  year      = {2022}
}