git clone https://github.com/nathanlct/trajectory_training.git
cd trajectory_training
conda env create -f environment.yml
conda activate trajectory
.py: These files are config files to set the hyperparameters for experiments.prompts/: Folder that contains different files for prompt engineering and experimentation.
Contains the core models and agents used in the project.
llm_agent.py: Defines the necessary classes to enable to use different languages models of your choice (Gemini, Groq, ChatGPT).
Contains the results of the user queries.
Contains the databases of situations that can be retrieved in the generation module.
json/: Contains.jsonfiles if needed in a human-readable format.pkl_db/: Contains.pklfiles for better storage.
evaluators/: All the content for the evaluation of the output in the training module.rag/: All the content for the retrieval part of the generation module.
File to run the entire framework.
In this work, we aim to provide a framework to tackle the issue of the black-box functioning of the autonomous vehicle controller. Our key contributions are the following:
- Development of a novel framework that translates complex RL decision-making into interpretable, human-centered explanations
- Validation of the approach using data from large-scale highway experiments with automated vehicles
- Advancement of explainable AI techniques in the critical domain of transportation systems
Our work is a continuation of the Vandertest 2021 ICRA work. The data of the Mega Vandertest we are using in this work can be found at: https://vanderbilt.app.box.com/s/z3nignz1cgm16dy56r1mqg9ycds40tkz
If you use the trajectory dataset in your published works please cite from this DOI:
For each situation, a dictionary is created with the following structure:
[
{
"current_situation": {"speed": 71.0, "headway": 35.0, "leader_speed": 70.5},
"action_explanation": "<observation>First observation<\/observation> <deduction>Second observation<\/deduction> <action>0.5<\/action>",
"action": 0.5,
"hypothetical_situation": "What happens if the headway suddenly decreases from 18m to 10m over the next 3 seconds?",
"situation_analysis": "<observation>First observation<\/observation> <deduction>Second observation<\/deduction> <prediction>The vehicle will brake in 2 seconds.<\/prediction>",
"embedding_model": "OpenAIEmbeddingModel()"
}
]python train.py --expname test --s3 --iters 200 --algorithm PPO --lr 3e-4 --n_epochs 10 --env_platoon 'av human*5'
Run python train.py - h for a description of all available args.
Note that a grid search can be ran over most args by specifying several values, for instance --lr 1e-4 5e-4 1e-5 --gamma 0.9 0.99 will run a total of 3 x 2 = 6 grid searches.
RL controller trained using train.py:
python simulate.py --cp_path checkpoints/.../n.zip --av_controller rl --gen_emissi
6BCB
ons --gen_metrics --platoon scenario1
Baseline controller, eg. IDM or FS:
python simulate.py --av_controller idm|fs --gen_emissions --gen_metrics --platoon scenario1
Run python simulate.py - h for a description of all available args.
To send a controller through the leaderboard pipeline, use --data_pipeline {your_name} {strategy / controller_name} 0.
Change 0 to 1 in case your run should be marked as the baseline on the leaderboard.
Steps to evaluate your custom controller
Define a vehicle class in env/vehicles.py (following the same format as IDMVehicle or FSVehicle for instance). Once created, go in env/simulation.py, import your vehicle class at the top of the file, then go to the add_vehicle method and add a mapping from your controller name to the vehicle class in the vehicle_class dict. Finally, you should be able to run python simulate.py --av_controller {your_controller_name}.
Running python simulate.py with the --gen_emissions flag will generate a .csv emission file from which you can extract metrics that are interesting to you. Some available scripts are:
python visualize/time_space_diagram.py {path_to_emissions.csv}to generate a time - space diagrampython visualize/platoon_mpg.py {path_to_emissions.csv}to generate a platoon MPG graph(valid if you ransimulate.pywith the--platoon scenario1flag)python visualize/render.py {path_to_emissions.csv}to render your controller in a Pygame window(not functional right now)
Additionally, a good number of plots and metrics are generated when running simulate.py with the --gen_metrics flag.
Note that the behavior of one controller may largely differ from trajectory to trajectory. simulate.py defaults to using a custom --seed {integer} so that the same trajectory is used across several runs, and when uploading to the leaderboard. Change the seed to run on a different trajectory, or use the --all_trajectories flag to run over all available trajectories. If using that flag, you can check out visualize/plots_from_emissions.py to plot some data across all trajectories, possibly comparing two controllers.