Quick Setup

Those instructions can be used to reproduce the same setup used in our field experiments.

Hardware Requirements

• NVIDIA Jetson TX2
• Intel Realsense D435
• DJI Matrice 100

Setup

Follow the following instructions to setup each robot computer: TX2 Basic Setup.

Pull the following Docker image on each robot: lajoiepy/doorslam-realsensed435-tx2-dji

Launch commands

• First you need to setup the aliases and high-level parameters.
  • MASK & OWN_IP: IP address info 192.168.MASK.OWN_IP.
  • MASTER: OWN_IP of the robot with the ROS master.
  • LOCAL_ROBOT_ID: Robot ID (from 0 to 2)
  • OTHER_ROBOT_ID: Neighbor ID (Assign 1 neighbor to each robot).
  • The other parameters do not need to be modified. Look to the paper for more details on those parameters.

• Make sure this script is at the root level to run the commands.

• A typical sequence:
  • Robot 0: camera, blabbermouth2, optimization, separators
  • Robot 1: camera, optimization, separators
  • Pilot the robots around the test area.
  • When you are ready for the distributed pose graph optimization execute trigger on either Robot 0 or 1.

• Reboot between each run.

• Commands description:
  • camera : Launches the Docker container and the onboard camera.
  • enter_bash : Opens a bash command line in the Docker container already running.
  • optimization : Launches the optimization process.
  • blabbermouth : Launches communication hub for 3 robots.
  • blabbermouth2 : Launches communication hub for 2 robots.
  • trigger : Triggers the distributed pose graph optimization process.
  • separators : Launches the inter-robot loop closures detection process.
  • record : Launches the data recording using rosbags.
  • docker_bash : Launches the Docker container and open a bash command line in it.
  • bag_sep : Launches the inter-robot loop closures detection process with a rosbag dataset (not using the camera).
  • bag_sep_params : Launches the inter-robot loop closures detection process with a rosbag dataset (not using the camera). Uses the parameters set here instead of the default ones.
  • cleancache : Cleans the TX2 cache. This usually has a positive effect on the computer performance. We recommend to run it often.

Good luck and have fun!

Building the Docker image for TX2

To build the Docker image yourself to replicate our exact setup, use this Dockerfile and instructions