Prunario

This repository contains the implementation of Prunario.

Installation

To setup Prunario, please refer INSTALL.md.

Overall Structure of the Implementation

The core functionalities of Prunario and their corresponding implementations are as follows:

• main.py: implements the main testing algorithm (Algorithm 1).
• fuzz/behavior/
  • behavior.py: implements the pattern-based transformation procedure (Section 3.1.1 Abstracting Simulation Results).
• fuzz/exp/
  • exp.py: implements redundancy checking procedures (Section 3.1.1, 3.2.3).
• fuzz/model/:
  • model.py: implements the training procedures for speed-prediction model (Algorithm 3).
  • utils.py: implements feature extraction procedures (Table 1).
• fuzz/simulator/
  • simulator.py: implements physical simulation procedures (line 11 in Algorithm 1).
  • simulator_abs.py: implements simulation prediction procedures (Algorithm 2).
  • planner_v.py: implements record prediction procedures (Algorithm 4, 5).

Run Prunario

• Please ensure that all procedures in INSTALL.md have been completed.

• Run the following command to run the testing loop:

  • We run wrapper.py instead of main.py to prevent the procedure from exiting due to a simulator error. The wrapper automatically restarts the simulator during execution.

  # Test Autoware in Town01 with Prunario
  python wrapper.py --path_seed seeds/seed_1 --path_log logs --gpu 0 --offscreen --town 1 --pruning
  # Test Autoware in Town03 with Prunario
  python wrapper.py --path_seed seeds/seed_3 --path_log logs --gpu 0 --offscreen --town 3 --pruning

• After running the command above, you can check the results under logs/.

• If you want to run Field and Basic (Section 5.2) without pruning, you can run the following command:

  # Test Autoware in Town01 with Field (static pruning)
  python wrapper.py --path_seed seeds/seed_1 --path_log logs --gpu 0 --offscreen --town 1 --naive
  # Test Autoware in Town01 with Basic
  python wrapper.py --path_seed seeds/seed_1 --path_log logs --gpu 0 --offscreen --town 1

Reproduce the results in Section 5

0. Download

Download all files from figshare and locate each files at the project root.

1. Table 2, 3, 4 and Figure 5

pip install -r requirements.txt
pip install data/carla-0.9.15-cp38-cp38-linux_x86_64.whl
mkdir dataset_path
tar -xzvf dataset.tar.gz -C dataset_path
export PYTHONPATH=.
python scripts/evaluate.py --path dataset_path

2. Results in the paragraph "Precision and Recall of Pruning" (Section 5.2)

pip install -r requirements.txt
pip install data/carla-0.9.15-cp38-cp38-linux_x86_64.whl
mkdir pnr_results
tar -xvzf pnr_results.tar.gz -C pnr_results
export PYTHONPATH=.
python scripts/precision_and_recall.py --path pnr_results