JETSON setup

Quick (actually very long) remainder how to setup a jetson nano

Hardware

---

What you actually need as material

• Jetson Nano Developer Kit

• Intel® Dual Band Wireless-AC 8265

• 5V 4A (4000mA) switching power supply (Any powersuply will do as is 5V, 4A, 5.5mm OD and 2.1mm ID)

• Micro SD (At least 32G)

• Standard Computer Jumper Caps

• A micro USB cable

Preparing the kit

---

You only need to put the kit together if you want you can follow the tutorials

• For the WIFI

• For use the barrel adapter

Step 1 - Flash micro SD card for the latest version

1. Download the image in the nvida site not use the sdk manager, that is currently not working right know

2. Insert a 32GB+ SD card into the desktop machine

3. Using Etcher unzip and select the image downloaded before

Step 2 - Power on and connect

1. Insert the configured SD card into the Jetson Nano module

2. Power on by plugging the powersuply to the jetson

3. Without a display plugged the jetson start in headless mode, you can connect to the jetson with the follow command this will guide you to install ubuntu and create the user and the wifi connection

   screen /dev/ttyACM0 115200

4. Connect to the jetson through ssh

Step 3 - Connect the Jetson to WiFi (If the wifi is not working)

1. Open a terminal

2. In the terminal, type the following command to list available WiFi networks, and find the ssid_name of your network.

   sudo nmcli device wifi list

3. Connect to the selected WiFi network

   It should be on the same network that you will be connect to jetson

   sudo nmcli device wifi connect <ssid_name> password <password>

4. Note down the WiFi IP returned by the following command. We'll call this jetson_ip_address

   ip route get 1.2.3.4 | awk '{print $7}'

5. Turn off the power safe mode to avoid losing the wifi connection

   sudo iw dev wlan0 set power_save off

Step 3.1 - Connect the Jetson to WiFi (In wap2 entreprise)

1. Follow the 2 first steps

2. Create WPA2 Entreprise connexion

   nmcli con add type wifi ifname wlan0 con-name <connexion-name> ssid <ssid_name>
   nmcli con edit id <connexion-name>
   set ipv4.method auto
   set 802-1x.eap peap
   set 802-1x.phase2-auth mschapv2
   set 802-1x.identity <user_name>
   set 802-1x.password <password>
   set wifi-sec.key-mgmt wpa-eap
   save
   activate

3. Continue as step 3

Step 4 - Connect to the Jetson over WiFi in ssh

1. Unplug the the Jetson Nano HDMI if you use a display or the micro usb if you use the headless mode

2. Open a new terminal in your main machine and connect to the jetson in ssh with the following command

   ssh <user_name>@<jetson_ip_address>

3. Sign in with the password <user_password>

Installing librealsense and ros

---

Step 1 - Realsense setup

Right now we are using the version 2.31

1. Add a swapfile to jetson

   git clone https://github.com/jetsonhacksnano/installSwapfile
   cd installSwapfile
   ./installSwapfile.sh
   cd ..

2. Add the ubuntu keyserver

   sudo apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv-key F42ED6FBAB17C654

3. Update the repo

   sudo apt update

4. Install dependencies

   sudo apt install -y wget git libssl-dev libusb-1.0-0-dev pkg-config libgtk-3-dev libglfw3-dev libgl1-mesa-dev libglu1-mesa-dev python3-dev cmake

5. Download librealsense

   sudo wget https://github.com/IntelRealSense/librealsense/archive/v2.31.0.tar.gz -O librealsense.tar.gz

6. Untar librealsense

   tar xvzf librealsense.tar.gz

7. Go to librealsense folder and create the build folder

   cd librealsense-2.31.0 && mkdir build && cd build

8. Run cmake build

    cmake ../ -DCMAKE_BUILD_TYPE=Release -DBUILD_PYTHON_BINDINGS:bool=true -DBUILD_WITH_CUDA:bool=true -DCMAKE_CUDA_COMPILER=/usr/local/cuda-10.0/bin/nvcc

9. Run Make with 4 cores

   make -j4

10. Install all the tools and libs

    sudo make install

11. Before using you need copy the udev rules for the realsense and restart udevadm

    cd ..
    sudo cp config/99-realsense-libusb.rules /etc/udev/rules.d/
    sudo udevadm control --reload-rules && udevadm trigger

Step 2 - ROS setup

We install ros melodic and the dependencies direclty from the repo

1. Add ros source to the deb list

   sudo sh -c 'echo "deb http://packages.ros.org/ros/ubuntu $(lsb_release -sc) main" > /etc/apt/sources.list.d/ros-latest.list'

2. Add the key to the key server

   sudo apt-key adv --keyserver 'hkp://keyserver.ubuntu.com:80' --recv-key C1CF6E31E6BADE8868B172B4F42ED6FBAB17C654 

3. Update apt list

   sudo apt-get update

4. Install ros desktop and rgbd launch

   sudo apt-get install -y ros-melodic-desktop ros-melodic-rgbd-launch

5. Install and run rosdep

   sudo apt-get install -y python-rosdep
   sudo rosdep init
   rosdep update

6. Add setup bash to bashrc and source bash rc

   echo "source /opt/ros/melodic/setup.bash" >> ~/.bashrc
   source ~/.bashrc

7. Install dependencies for ros

   sudo apt-get install python-rosinstall python-rosinstall-generator python-wstool build-essential python-catkin-tools  -y

8. Install dependecies for python3

   sudo apt-get -y install python3-pip python3-yaml
   sudo pip3 install rospkg catkin_pkg

Step 3 - Realsense ROS setup

The current master support the version 2.31 of realsence

1. Create the workspace for realsense ROS

   mkdir -p ~/catkin_ws_rs/src
   cd ~/catkin_ws_rs
   catkin init
   cd src

2. Clone realsense ros and ddynamic reconfigure

   git clone https://github.com/pal-robotics/ddynamic_reconfigure
   git clone https://github.com/IntelRealSense/realsense-ros.git

3. Checkout the branch 2.2.12 to go with our version and change CMakelist to works also with the 2.2.12 version

   cd realsense-ros/
   git checkout 2.2.12
   sed -i 's/find_package(realsense2 2.25.0)/find_package(realsense2 2.28.0)/' realsense2_camera/CMakeLists.txt 
   cd ../..

4. Use catkin to make the package

   catkin config -DCATKIN_ENABLE_TESTING=False -DCMAKE_BUILD_TYPE=Release
   catkin build

5. Add the workspace to the PATH and the .bashrc file

   echo "source /home/jetson/catkin_ws_rs/devel/setup.bash --extend" >> ~/.bashrc
   source ~/.bashrc

Intalling mavros

---

1. Create the workspace: unneeded if you already have workspace

   mkdir -p ~/catkin_ws_mavros/src
   cd ~/catkin_ws_mavros
   catkin init
   wstool init src

2. Install MAVLink

   We can use the melodic reference for all ROS distros as it's not distro-specific and is up to date --rosdistro melodic this example use the master

   rosinstall_generator --upstream-development mavlink | tee /tmp/mavros.rosinstall

3. Install MAVROS: get source (upstream - released)

   Install latest source

   rosinstall_generator --upstream-development mavros | tee -a /tmp/mavros.rosinstall

4. Create workspace & deps

   wstool merge -t src /tmp/mavros.rosinstall
   wstool update -t src -j4
   rosdep install --from-paths src --ignore-src -y

5. Install GeographicLib datasets:

   sudo ./src/mavros/mavros/scripts/install_geographiclib_datasets.sh

6. Build source

   catkin build

7. Add the workspace to the PATH and the .bashrc file

   echo "source /home/jetson/catkin_ws_mavros/devel/setup.bash --extend" >> ~/.bashrc
   source ~/.bashrc

8. Set the rights to the ttyTHS in the udev rules

   sudo sh -c 'echo "KERNEL==\"ttyTHS*\", MODE=\"0666\""  > /etc/udev/rules.d/55-jetsonserial.rules'
   systemctl stop nvgetty
   systemctl disable nvgetty
   udevadm trigger (or reboot)

Intalling tensorflow

---

Right now we are using the version 1.14

1. Installing dependencies

   sudo apt-get install -y libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev zip libjpeg8-dev

2. Installing pip and python dependencies

   sudo apt-get install -y python3-pip 
   sudo pip3 install -U pip testresources setuptools
   sudo pip3 install -U numpy==1.16.1 future==0.17.1 mock==3.0.5 h5py==2.9.0 keras_preprocessing==1.0.5 keras_applications==1.0.8 gast==0.2.2 enum34 futures protobuf
   

3. Installing official tensorflow for nano, build by Nvidia

   $ sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v43 tensorflow-gpu