ec-jrc / jrc-fsm

Regulation 157 Safety Models

This document describes how to use the following repository to investigate the behavior of the proposed safety models for UNECE R157.

Three reference scenarios are implemented in the current formulation:

1. 'cut_in';
2. 'cut_out';
3. 'car_following'.

Four models are here discussed:

1. 'FSM' : Fuzzy Safety Model [1];
2. 'RSS' : Responsibility Sensitive Safety [2];
3. 'CC_human_driver' : [3];
4. 'Reg157' : [4];

The python script 'safety_check_runner.py' provides the possibility of selecting three types of analyses:

1. 'one_case' : enables selecting one concrete scenario, one model, and visually inspecting the result of the simulation;
2. 'comparison' : aims at a systematic investigation of the safety models on a selection of logical scenario and then stores data for later processing;
3. 'post_processing' : provides the possibility of visually inspect the results of the previously executed 'comparison' scenario.

The 'safety_check_runner.py' can be launched either via command line providing the minimum number of arguments or called from another python script as shown in the 'example.py' file.

One Case

The 'one_case' analysis is a single simulation which involves a concrete scenario with a specified safety model.

An example command line instruction to execute a 'cut_in' scenario is:

python safety_check_runner.py one_case cut_in

The default model for the 'one_case' analysis is FMS, however other models can be executed too as:

python safety_check_runner.py one_case cut_in model=RSS

The concrete scenarios can be modified via passing optional parameters.
In particular, the following parameters can be adjusted on the fly:

• the initial velocity of the simulation can be set via passing initial_speed optional command in (km/h):

python safety_check_runner.py one_case cut_in model=RSS initial_speed=50

• the leader deceleration of car-following scenario can be adjusted (in m/s²) as:

python safety_check_runner.py one_case car_following model=Reg157 deceleration=3

• the obstacle speed in the cut-in scenario can be updated (in km/h, default=) via:

python safety_check_runner.py one_case cut_in model=CC_human_driver obstacle_speed=20

• the lateral speed of cut-out and cut-in maneuvers can be adjusted (in m/s, default=-1.) as:

python safety_check_runner.py one_case cut_in model=CC_human_driver lateral_speed=-2.

• the front distance of cut-out and cut-in maneuvers can be adjusted (in m, default=50) as:

python safety_check_runner.py one_case cut_in model=FSM front_distance=30

Comparison

The comparison cases loop over the safety models for a range concrete scenarios belonging to the same class.

Three alternatives are possible corresponding to the

• cut-in scenario:

python safety_check_runner.py comparison cut_in

• cut-out scenario

python safety_check_runner.py comparison cut_out

• car-following scenario

python safety_check_runner.py comparison car_following

No optional parameters can be provided in the 'comparison' analysis.

Post processing

Eventually, a post processing analysis can be run to save/visualize the results of the comparisons.

• cut-in post-processing:

python safety_check_runner.py post_processing cut_in

- a detailed analysis on TTC is possible for the cut_in case by passing a safety model specification  
 
    `python safety_check_runner.py post_processing cut_in model=RSS` 

- a detailed analysis on FSM criticality is possible for the cut_in case by passing the FSM model specification  
 
    `python safety_check_runner.py post_processing cut_in model=FSM` 

• cut-out scenario

python safety_check_runner.py post_processing cut_out

• car-following scenario

python safety_check_runner.py post_processing car_following

Finally, a 'save_image' boolean can be passed to automatically store the image on the hard drive instead of the graphical visualization only.

python safety_check_runner.py post_processing car_following save_image=True

Dependencies

The following python packages are required to run the models:

• pandas (New BSD License, 3-clause);
• numpy (New BSD License, 3-clause);
• numba (BSD 2-clause);
• scipy (BSD);
• matplotlib (BSD).

References

[1] Mattas, Konstantinos, et al. "Fuzzy Surrogate Safety Metrics for real-time assessment of rear-end collision risk. A study based on empirical observations." Accident Analysis & Prevention 148 (2020): 105794.

[2] Shalev-Shwartz, Shai, Shaked Shammah, and Amnon Shashua. "On a formal model of safe and scalable self-driving cars." arXiv preprint arXiv:1708.06374 (2017).

[3] UNECE Reg 157 Annex 4 - Appendix 3

[4] UNECE Reg 157 Paragraph 5.2.5.2.

jrc-fsm's Issues

I have a question about the "-0.72" when using CC-human model to analysis the cut-in scenario

In the code "model.py", there is a "-0.72" in Line 50.
I checked the UNECE Reg 157 Annex 4 - Appendix 3 and there is a figure also shows "0.72m"

However, I couldn't figure out what the meaning of it. Could you help me explain its meaning？Thanks