Skip to content

Latest commit

 

History

History
104 lines (84 loc) · 5.71 KB

setup.md

File metadata and controls

104 lines (84 loc) · 5.71 KB
Raw

Setup Instructions

Radar (Windows) Computer

  1. If using RDP to control the data collection system, set up a RDP connection on the Windows laptop.

    1. Open the Remote Desktop Connection program (should be a default program on Windows).
    2. Enter the computer name that you've set, i.e. dart-radar.local, and your username.
    3. (Optional) Create a shortcut: click Show Options, then Save As....

  2. If manual time synchronization is required, set up the Windows computer as a NTP server.

    • You can also use the linux computer as a NTP server or a common external NTP server (preferred). Only follow these instructions if that is not possible.
    • See full instructions here.

  3. Install mmWave studio & dependencies. See the DCA1000EVM Quick Start Guide for full instructions.

    • Make sure to set a static IP address (192.168.30.33) for the network interface used as shown in step 3.
    • Make sure to install the matlab runtime engine noted in step 4.
    • You may need to install the mmWave SDK.
    • Make sure that all 6 COM ports are detected as shown in Step 6 / Figure 4. If the XDS110 Class ports are not detected, see the note about installing EMUPACK.
    • NOTE: you may need to disable windows firewall, which might block the ports used by the radar by default.
    • NOTE: keep a copy of the mmWave installer. If the radar computer loses power while mmWave studio is running, it may become corrupted in a subtle way and require reinstallation (the symptoms we observed were a "null reference" error when trying to start data collection (but not initialize the radar/capture card), and mmWave studio failing to read the FPGA version).

  4. Clone this repository and install dependencies.

    git clone [email protected]:thetianshuhuang/rover
    • NOTE: submodules (i.e. recursive cloning) are not used by the radar data collection system.
    • Install Python >3.8, and pip install -r radar/requirements.txt.

  5. Configure the lua scripts.

    • Note which COM port is labeled "XDS110 Class Application/User UART". You can verify that you have selected the correct one using mmWave studio:
      1. Open mmWave studio.
      2. Click the Set button under Board Control / Reset Control / Reset. Wait for the board to finish resetting.
      3. Select the COM port in the Board Control / RS232 Operations dropdown.
      4. Click Connect. The RS232 Connectivity Status field should turn to Connected.
    • Copy sample configuration rover/radar/_config.json to rover/radar/config.json, and update it with your configuration parameters.
      • Make sure to set the com entry with the COM port number noted above.
      • Unless you have a non-standard mmWave Studio installation, none of the other fields need to be changed.
    • Build the lua scripts: 
      python build.py

LIDAR (Ubuntu 20.04) Computer

  1. If using SSH to control the data collection system, make sure to install/enable SSH.

    • A virtual terminal client like screen, tmux, zellij is also helpful. We use screen in these instructions (sudo apt-get -y install screen).

  2. Synchronize time if required (sudo apt-get install -y ntpdate if needed):

    sudo ntpdate dart-radar.local

  3. Set Link-Local only. In the GUI:

    1. Go to Settings/Network. Click on the ''settings'' button next to the wired connection.
    2. In the IPv4 tab, set IPv4 Method to Link-Local Only.
    3. In the IPv6 tab, set IPv6 Method to Link-Local Only.
    4. Disable and re-enable the wired connection.

  4. Install ROS Noetic

    • NOTE: make sure you activate the ROS environment on each terminal you use to run subsequent ROS commands: 
      source /opt/ros/noetic/setup.bash

  5. Clone this repository. Make sure to clone all subrepositories!

    git clone [email protected]:thetianshuhuang/rover --recursive
    • Use git submodule update --init --recursive if you forgot to git clone --recursive.

  6. Install ROS dependencies.

    • Install dependencies for ouster-ros.
    • Install python3-catkin-tools: 
      sudo apt-get -y install python3-catkin-tools
    • Follow manual installation instructions in lidar/catkin_ws/src/xsens_ros_mti_driver: 
      pushd src/xsens_ros_mti_driver/lib/xspublic && make && popd

  7. Build:

    source /opt/ros/noetic/setup.bash
catkin build --cmake-args -DCMAKE_BUILD_TYPE=Release
    • Use catkin clean to delete all built files if a rebuild is needed.
    • --cmake-args -DCMAKE_BUILD_TYPE=Release is needed for ouster-ros.

Data Processing Computer

Install the following on a fast computer, ideally with both good single-core CPU performance and a GPU.

Cartographer: You're on your own for this one.

  1. ROS Noetic is required.
  2. Cartographer ROS placed inside the currently sourced catkin_ws.

Python: The python environment for DART is a superset of rover, so you can use that if you have one set up.

  1. Ensure that you have python (>=3.8) and CUDA (>=11.8).
  2. Install jax.
  3. Install libhdf5: sudo apt-get -y install libhdf5-dev
  4. Install python dependencies: pip install -r processing/requirements.txt