DonkeyCar Performance Measurement

About The Project

Introduction

This repository provides scripts for computing Cross Track Error (CTE) from video data captured while running trained models on the DonkeyCar platform. The CTE is calculated by analyzing vehicle trajectory relative to the track centerline using overhead video recorded from a ceiling-mounted, downward-facing camera.

Project Layout

src
 ├── main.py 
 ├── Mean_and_std_dev.py 
 └── draw_plot.py

Dependencies

• opencv
• numpy
• matplotlib
• scipy

Usage

Instructions

1. Clip the video.

2. Organize the video files:

   • Create a video folder under the project root directory.
   • Place the original and clipped versions of the video in the video folder.
     • Video format must be .mp4.
     • Name the clipped version as: original_name_cliped.mp4.

3. Run the main script:

   python3 src/main.py your_video_file_name

• Do not include the .mp4 extension in your_video_file_name.
• The output will be saved in a new subfolder under the data directory.
• The results will include:
  • An annotated video.
  • A CSV file containing CTE (Cross-Track Error) values for each frame.

4. Compute mean and standard deviation of CTEs:

python3 src/Mean_and_std_dev.py subfolder_name

5. Visualize CTE over time:

python3 src/draw_plot.py subfolder_name

This will generate a real-time CTE plot aligned with the annotated video.

License

Distributed under the MIT License. See LICENSE for more information.

Authors

Tyler Ruble, Zhongzheng R. Zhang, Giancarlo Vidal