This is the project repo for the final project of the Udacity Self-Driving Car Nanodegree: Programming a Real Self-Driving Car. For more information about the project, see the project introduction here.
- Georgios Fagogenis, georgios.fagogenis@tttech-auto.com
- Matias Fernando Pavez Bahamondes, matias.pavez-bahamondes@tttech-auto.com (Team Lead)
- Natalia Estefanía Jurado Espín, natalia.jurado@tttech-auto.com
- Installation: Local/Docker setup, simulator, and dependencies.
- Testing: How to test on the simulator and the real car data.
- Implementation: Design and implementation details.
- Results.
- Acknowledgements.
Compile the project locally or on the docker container:
cd ros
source /opt/ros/kinetic/setup.bash
catkin_make
source devel/setup.sh
- Run the simulator locally: The camera must be enabled and the manual mode disabled.
- Launch styx.launch locally or on the docker container.
cd ros
roslaunch launch/styx.launch
- Run the Rosbag file that was recorded on the Udacity self-driving car:
- Launch the project in your project in site mode
rosbag play -l traffic_light_bag_file/traffic_light_training.bag
cd ros
roslaunch launch/site.launch
- TODO: About decceleration profile for the trajectory.
- TODO: How the PID for throttle was tuned.
- TODO: Insights on the yaw controller.
- TODO: How to deal with stop and go scenarios.
Labeled data, and training scripts for the traffic light detector were provided by user vatsl on the repository TrafficLight_Detection-TensorFlowAPI.
A different container matching the exact dependencies as in the real car was used for training the model. Instructions can be found in the respective training section.
- TODO: About the dataset.
- TODO: About the model.
- TODO: About the results for simulation.
- TODO: About the results for real car data.
- TODO: Add images/video.
- TODO: (Improvement - Corner case): About how dealing with yellow lights would improve the stop.
- Most of the implementation for ROS nodes and controllers was obtained from Udacity classroom lessons.
- Labeled data, and training scripts for the traffic light detector were provided by user vatsl on the repository TrafficLight_Detection-TensorFlowAPI.