/AVARS

AVARS - Alleviating Unexpected Urban Road Traffic Congestion using UAVs

Primary LanguagePythonMIT LicenseMIT

AVARS - Alleviating Unexpected Urban Road Traffic Congestion using UAVs

AVARS system flow

The experiments are conducted on a simulator platform SUMO. The model design and implementation are based on Flow, which configures with RLlib, an open-source library for reinforcement learning(RL) and ease to execute efficient and scalable RL experiments due to integrated with Ray.

Installation

Install Anaconda

It is highly recommended to install Anaconda that is convenient to set up a specific environment for Flow and its dependencies.

Install FLOW

Download this project, which covers the whole framework of Flow and my model implementation based on Flow.

git clone git@github.com:Guojyjy/AVARS.git

Create a conda environment

First, cd ~/AVARS/flow to the directory, then enter the following commands and run the relevant scripts to create the project environment and activate it, install Flow and its dependencies:

conda env create -f environment.yml
conda activate flow
python setup.py develop # if the conda install fails, try the next command to install the requirements using pip
pip install -e . # install flow within the environment

If you have trouble installing FLOW, the official Flow documentation provides more installation details: Local installation of Flow.

In addition, because the definition of $SUMO_HOME within the installation process of SUMO would cause an error in the installation of Flow, please confirm the installation of Flow first.

Install SUMO

It is highly recommended to use the official installation methods from Downloads-SUMO documentation.

The instructions covered in Installing Flow and SUMO from Flow documentation is outdated.

# run the following commands to check the version/location information or load SUMO GUI
which sumo
sumo --version
sumo-gui

Troubleshooting

  1. See output like Warning: Cannot find local schema '../sumo/data/xsd/types_file.xsd', will try website lookup.
    • Set $SUMO_HOME to $../sumo instead of $../sumo/bin
  2. Error like ModuleNotFoundError: No module named 'flow', ImportError: No module named flow.subpackage
    • pip install -e . to install flow within the environment, mentioned at Install FLOW
    • Issue on inconsistent version of python required in the environment
      • Linux version of SUMO contains python in /sumo/bin/ which may cause the error.
      • which python to check the current used
      • echo $PATH to check the order of the directories in the path variable to look for python
      • Add export PATH=/../anaconda3/env/flow/bin:$PATH in the file ~/.bashrc
      • Restart the terminal, update the configuration throughsource ~/.bashrc

Experiments

Enter the project in the specific environment:

cd ~/AVARS/flow
conda activate flow

Scenario settings

Extracted from the open data in maxime-gueriau / ITSC2020_CAV_impact to simulate the real-world traffic in Dublin city.

Road network

A subnet of Dublin city center road network around the River Liffey, covering approximately 1 square kilometer, is illustrated as the traffic environment in the system flow diagram above.

Traffic

Traffic generation lasts 45 minutes with 1168 vehicles.

Compared scenarios

AVARS

# locate in ~/AVARS/flow
python examples/train_ppo.py AVARS_UAV --num_steps 150
python examples/train_dqn.py AVARS_UAV_dqn --num_steps 150
  • train_ppo.py and train_dqn.py include DRL algorithm assigned and parameter settings
    • num_gpu, num_worker specify the computation resource used in the DRL training
  • AVARS_UAV and AVARS_UAV_dqn both correspond to the modules of flow/examples/exp_configs/rl/multiagent, including all setting of road network and DRL agent
  • "--num_steps" is the termination condition for DRL training, optional
  • The SUMO files of Dublin scenario locate in AVARS/scenarios/UAV
  • System design in flow/flow/env/multiagent/sumo_template.py
    • UAVEnvAVARS
    • UAVEnvIntelliLight

Original

cd ~/AVARS/SCATS
python Imple_SCATS.py --scen center10 --nogui --noscats --run_num 18

Congestion

cd ~/AVARS/SCATS
python Imple_SCATS.py --scen center10_closing --nogui --noscats --run_num 18

SCATS[1]

cd ~/AVARS/SCATS
python Imple_SCATS.py --scen center10_SCATS --nogui --run_num 18

IntelliLight[2]

# locate in ~/AVARS/flow
python examples/train_ppo.py IntelliLight_UAV_ppo --num_steps 150
python examples/train_dqn.py IntelliLight_UAV --num_steps 150

Evaluation

  • evaluation/outputFilesProcessing.py filters the output files in AVARS/output

    • delete outdated, incomplete, and initial (size < 4kB ) files

    • recover the format of xml files

  • evaluation/getResults.py gets traffic statistic

    • Travel time
    • Fuel consumption
    • CO2 emissions

    python getResults.py --scen center10 center10_closing center10_SCATS ...

  • evaluation/draw_reward_episode.py generates lineplot of average episode reward for AVARS-controlled TLC, based on progress.csv in each training stored at ray_results/

  • evaluation/draw_runningveh.py uses the output from sumo-gui when running Imple_SCATS.py or flow/flow/visualize/visualizer_rllib.py

Citing

References

[1] P. Lowrie, “SCATS, sydney co-ordinated adaptive traffic system: A traffic responsive method of controlling urban traffic,” 1990.

[2] H. Wei, G. Zheng, H. Yao, and Z. Li, “Intellilight: A reinforcement learning approach for intelligent traffic light control,” in Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2018, pp. 2496–2505.