harmoniousharmony / system_integration Goto Github PK

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.

###Team

• Be sure that your workstation is running Ubuntu 16.04 Xenial Xerus or Ubuntu 14.04 Trusty Tahir. Ubuntu downloads can be found here.

• If using a Virtual Machine to install Ubuntu, use the following configuration as minimum:

  • 2 CPU
  • 2 GB system memory
  • 25 GB of free hard drive space

  The Udacity provided virtual machine has ROS and Dataspeed DBW already installed, so you can skip the next two steps if you are using this.

• Follow these instructions to install ROS

  • ROS Kinetic if you have Ubuntu 16.04.
  • ROS Indigo if you have Ubuntu 14.04.

• Dataspeed DBW

  • Use this option to install the SDK on a workstation that already has ROS installed: One Line SDK Install (binary)

• Download the Udacity Simulator.

1. Clone the project repository

git clone https://github.com/srikantrao/System_Integration.git

2. Install python dependencies

cd CarND-Capstone
pip install -r requirements.txt

3. Make and run

cd ros
./clean.sh
source devel/setup.sh

4. Running different launch files for different scenarios

If you want to run the simulator, run the following command

roslaunch launch/sim.launch

If you want to run it on the car, run the following command

roslaunch launch/site.launch

If you want to see a visualization of the detector working

roslaunch launch/site_visualization.launch

5. Run the simulator

1. Download training bag that was recorded on the Udacity self-driving car (a bag demonstraing the correct predictions in autonomous mode can be found here)
2. Unzip the file

unzip traffic_light_bag_files.zip

3. Play the bag file

rosbag play -l traffic_light_bag_files/loop_with_traffic_light.bag

4. Launch your project in site mode

cd CarND-Capstone/ros
roslaunch launch/site.launch

5. Confirm that traffic light detection works on real life images

You can confirm that the detector works on real life images by running the rosbag

cd CarND-Capstone/ros
roslaunch launch/site_visualization.launch

Open another terminal and go to CarND-Capstone/ros

source devel/setup.bash
rosbag play -l traffic_light_bag_files/loop_with_traffic_light.bag

Open another terminal and go to CarND-Capstone/ros

source devel/setupbash
rosrun tools rosbagVisual.py

We have explored multiple options to detect traffic lights: firs t one was building our own Deep Convolutional Neural Network (CNN) and using pre-trained models.

What we noticed that we need a more data to train our CNN models to have a good model. Due to the limmitted training data and training time constraints we ended up using pre-trained object detection model (Fast Region-based Convolutional Network ). Here is the model link: https://github.com/tensorflow/models/tree/master/research/object_detection

A jupyter notbook was created to test the model accuracy and here are the some outputs. Please refer to the notebook, ~/classifier/faster-R-CNN/object_detection.ipynb, for more details.

The twist controller node was implemented using a PID controller for brake/throttle.

At a high level, the throttle value was given by

throttle = Kp * vel_err + Kd * vel_err_delta + Ki * vel_err_integral

If the throttle was negative, it was considered a brake value. The steering angle was determined based on code provided in yaw_controller.py

For the same radius

v_obs * r = w_obs and v_ctrl * r = w_ctrl

w_ctrl = w_obs * v_ctrl / v_obs

A low pass filter usign exponentially weighted averages was used to smoothen the steering output result. The steering angle output is effectively a moving window average of the last 20 values for simulation and last 5 values for the site.