Year: 2017

We trained a CNN model to predict steering angles from the visual input. Here is an example of our accomplishments.

We used the same CNN model on a different track to show the flexibility of this approach. The followings show as successful driving on a different track from one where the model was trained.

Source code can be found at https://github.com/jrkwon/mir_torcs

Unlike TORCS, Euro Truck Simulator 2 is a commercial product. The regular price is around $20. But as in the article below, an open source project, europilot (https://github.com/marshq/europilot) helps us to capture data for training.

• Create self-driving trucks inside Euro Truck Simulator 2

https://github.com/philbort/awesome-self-driving-cars

We can use /dev/lidar instead of /dev/ttyUSB#. Refer the followings.

• http://hintshop.ludvig.co.nz/show/persistent-names-usb-serial-devices/
• http://www.joakimlinde.se/microcontrollers/arduino/avr/udev.php

I used following references.

• Installation of opencv-3.1.0 on Jetson TX1 [here]
• Installing OpenCV 3.1.0 [here]
• minhoolee/install-opencv-3.1.0.sh [here]
• dselivanov/scripts-ubuntu-debian [here]

Uninstall the current OpenCV

• $ sudo rm -rf /usr/local/include/opencv
• $ sudo rm -rf /usr/local/include/opencv2
• $ suro rm -rf /usr/local/lib/python2.7/dist-packages/cv2.*
• $ suro rm -rf /usr/local/lib/python3.5/dist-packages/cv2.*
• $ sudo rm /usr/local/lib/libopencv_*

You need to change the Python version numbers if you are using different ones.

Keep Ubuntu up to date

• $ sudo apt-get -y update
• $ sudo apt-get -y upgrade
• $ sudo apt-get -y dist-upgrade
• $ sudo apt-get -y autoremove

Install development libraries

• # Add universal repository to Ubuntu
• $ sudo apt-add-repository universe
• $ sudo apt-get update
• # Some general development libraries
• $ sudo apt-get install -y build-essential make cmake cmake-curses-gui g++
• # libav video input/output development libraries
• $ sudo apt-get install -y libavformat-dev libavutil-dev libswscale-dev
• # Video4Linux camera development libraries
• $ sudo apt-get install -y libv4l-dev
• # Eigen3 math development libraries
• $ sudo apt-get install -y libeigen3-dev
• # OpenGL development libraries (to allow creating graphical windows)
• $ sudo apt-get install -y libglew1.6-dev
• # GTK development libraries (to allow creating graphical windows)
• $ sudo apt-get install -y libgtk2.0-dev
• # Install unzip and wget
• $ sudo apt-get install -y unzip wget

Download and install OpenCV

• # Download
• $ cd ~/Downloads
• $ wget https://github.com/opencv/opencv/archive/3.1.0.zip
• # Unzip
• $ unzip 3.1.0.zip
• $ rm 3.1.0.zip
• # Prepare for building
• $ cd opencv-3.1.0
• $ mkdir build $ cd build
• # Building
• $ cmake -DWITH_CUDA=ON -DCUDA_ARCH_BIN=”5.3″ -DCUDA_ARCH_PTX=”” -DBUILD_TESTS=OFF -DBUILD_PERF_TESTS=OFF -DCUDA_FAST_MATH=ON ..
• $ make -j4
• $ sudo make install
• # Setup
• $ sudo /bin/bash -c ‘echo “/usr/local/lib” > /etc/ld.so.conf.d/opencv.conf’
• $ sudo ldconfig

When an error message is shown from a python code using cv2 (opencv), the message is most likely misleading. The message is saying nothing but something is wrong. All the details in the message are not what you need to do to fix the problem.

I had issues with VideoCapture and imshow. The error messages look explanatory but they are not. All problems were stemmed from a python wrapper package that I installed through pip.

$pip uninstall opencv-python