- Introduction
- Features
- Installation
- Usage
- Future Plans
- Improved Bird's-eye View
- Contributing
- License
- Acknowledgments
This project simulates autonomous car systems using a game environment called City Car Gaming, known for its realistic mechanics and graphics, sufficient to track objects and simulate real-world scenarios.
The system tracks various objects in the game environment using yolov8n model.
Transforms the front camera view into a bird's-eye view to map surroundings and predict the next path of objects. Currently, it is able to:
- Show a bird's-eye view for close distance objects
- Warn about possible collisions based on short-range predictions but only in straight road for now
Detects road lines using a sliding window technique, which assists in maintaining lane integrity.
git clone https://github.com/elymsyr/autonomous-systems-simulation.git
pip install -r requirements.txt
python run_simulation.py
Enhancing the bird's-eye view system to generate a more detailed environmental map and implement car pose detection. The goal is to provide accurate 3D terrain detection and car orientation.
Improving object path prediction to handle long-range scenarios and diverse object movement. The goal is to develop more accurate trajectory predictions for moving objects.
This system will provide self-driving capabilities by creating a path for the car to follow autonomously. The plan includes path planning algorithms and integration with the object detection module.
Introducing a system to prevent collisions using real-time object detection and prediction. Future implementations might include autobraking or steering interventions.
Contributions are welcome. Please check the issues tab and submit pull requests. See .github
See the LICENSE file for details.
I would like to thank the following resources and projects that inspired or contributed to features:
@article{gosala21bev,
author={Gosala, Nikhil and Valada, Abhinav},
journal={IEEE Robotics and Automation Letters},
title={Bird’s-Eye-View Panoptic Segmentation Using Monocular Frontal View Images},
year={2022},
volume={7},
number={2},
pages={1968-1975},
doi={10.1109/LRA.2022.3142418}}