Helium

A drone synthetic environment development project based on Ardupilot, Gazebo & ROS.

Learning the Basics

https://github.com/yanhwee/ardupilot-gazebo-ros-guide

Installation

Installation Script

From the terminal:
wget -O - https://raw.githubusercontent.com/yanhwee/helium/master/docs/install.bash | bash

This installation script is modified based on the installation from https://github.com/yanhwee/ardupilot-gazebo-ros-guide. Refer to it for more details. It installs the following prerequisites and setups this repository. Note that this repository is a catkin package. Please restart after the installation.

Prerequisites

Main SITL Stack

• ArduPilot (Modified)
• ArduPilot Gazebo Plugin (Modified)
• Gazebo 9
• ROS Melodic
• MAVROS

Supporting Stack

• Catkin Workspace & Helium Package
• QGroundControl (QGC)
• Google Protobufs
  • Generate Python Bindings for Gazebo Protobuf Messages
• Python Libraries
  • pymavlink
  • protobuf

Software Architecture Overview

Starting the SITL Stack

Manually

1. Open four terminals in the home directory (Tip: Ctrl-Alt-T + Ctrl-Shift-T x4)
2. In each:
   1. roslaunch Gazebo world
      roslaunch helium hills_lidar.launch
   2. Start ArduPilot SITL
      sim_vehicle.py -v ArduCopter -f gazebo-iris
      • Will need to compile (for first time)
      • Wait for APM: EKF2 IMU1 tilt alignment complete before continuing to step 3
      • Initialisation is completed when APM: EKF2 IMU1 is using GPS
        • Then you can start sending MAVLink commands & interface with MAVProxy
   3. Start MAVROS
      roslaunch helium apm.launch
   4. Start QGroundControl
      ./QGroundControl.AppImage

Automatically

Note: Run it via the gnome terminal.

1. Start: bash run.bash
2. Stop: ctrl-c

Repository Guide

1. docs
   • Contains documentation, diagrams, installation script, code examples
2. launch
   • Contains ROS launch files
   • apm.launch starts MAVROS
   • The rest starts Gazebo worlds within ROS
3. models
   • Contains Gazebo models
4. worlds
   • Contains Gazebo worlds
5. src
   • Contains SITL Code
   • mavdrone.py
     • Module for communicating with the drone via MAVLink
     • Uses pymavlink beneath to handle MAVLink messages.
   • gazebo.py
     • Module for retrieving Gazebo topic messages.
     • Uses protobuf to parse gztopic text formatted messages.
   • time_mission.py
     • Code example for uploading and timing drone missions
   • multi_drone.py
     • Code example for multi_drone simulation
6. setup.bash
   • Is sourced in .bashrc
   • Setups Gazebo world and model paths

Tips

1. To change directory to helium quickly:
   • roscd helium

Bugs Found

1. During a mission land command, if terrain following is enabled, the drone will keep flying higher.

Features Implemented for ArduCopter SITL

For ArduCopter Version: 4.0.4-dev

In Forked Repositories:

• ArduPilot
  https://github.com/yanhwee/ardupilot

• ArduPilot Gazebo Plugin (khancyr)
  https://github.com/yanhwee/ardupilot_gazebo

1. Terrain Following

1. References

   1. https://discuss.ardupilot.org/t/gazebo-with-lidar/24717/6

2. Psuedo-Diagram
   

3. Walk-through

   1. SIM_Gazebo

      1. Modify the FDM Packet to receive the rangefinder value.
      2. Then, moidfy the function that updates the range value (in SIM_Gazebo) using the received rangefinder value.

   2. ArduPilotPlugin.hh/.cc

      1. Modify the FDM Packet to take in the rangefinder value.
      2. Use Gazebo transport node to read laser topic messages published by laser ray model.

   3. Gazebo

      1. Create a model file for Lidar.
      2. Use ray sensors with one laser.
      3. Add this lidar model to iris_with_ardupilot, using links and joints.
         • Laser topic messages is automatically published.

4. ArduPilot Parameters

   1. References
      1. https://ardupilot.org/copter/docs/terrain-following.html
   2. By QGroundControl or MAVProxy
      1. TERRAIN_ENABLE = 1 (Default)

5. Test

   1. QGroundControl

      1. Select MAVLink Inspector
      2. Find Distance Sensor
      3. Plot, 60 secs (optional)

   2. MAVProxy (Alternatively, less preferred way)

      1. https://ardupilot.org/dev/docs/using-sitl-for-ardupilot-testing.html#adding-a-virtual-rangefinder
      2. Don't param set anything
      3. Just load the graph

2. Object Avoidance

1. References

   1. ArduPilot/ardupilot#5608
   2. https://answers.gazebosim.org//question/18881/publish-a-gazebo-message-into-a-ros-topic/

2. Walk-through

   1. Gazebo

      1. Create model file for 360 lidar.
         1. Ensure the lasers are setup in a "circular" manner
            • Actually clockwise or not it doesn't matter
              • Rather, ArduPilot parameter, PRX_ORIENT, controls the orientation of the proximity sensors
            • Just ensure that min_angle starts from 0
            • Don't overlap the first and last laser

            <samples>n</samples>
            <min_angle>0</min_angle>
            <max_angle>-x<max_angle>
            <!-- where x = 2pi / n * (n - 1) -->
            

      2. But this time, publish laser messages via MAVROS.
         1. Use a Gazebo Plugin, "libgazebo_ros_laser.so".
         2. Set publish topic target directly to "/mavros/obstacle/send".
            • It is a mavros rostopic that takes in Gazebo laser messages.
      3. Put lidar model on top of the iris_with_ardupilot using links and joints.

   2. Catkin Workspace

      1. Use roslaunch to run Gazebo. Create a launch file for that.
         • This is for "libgazebo_ros_laser.so", to be able to publish laser messages to rostopic.

3. ArduPilot Parameters

   1. References
      1. https://ardupilot.org/copter/docs/common-simple-object-avoidance.html
      2. https://ardupilot.org/copter/docs/common-oa-bendyruler.html
   2. By QGroundControl or MAVProxy
      1. AVOID_ENABLE = 7 (or any values that includes "Proximity Sensors")
      2. PRX_TYPE = 2 (MAVLink)
      3. PRX_ORIENT = 1 (Upside Down)
      4. OA_TYPE = 1 (BendyRuler)

4. Test

   1. Via Terminal
      1. rostopic echo /mavros/obstacle/send
   2. QGroundControl
      1. Ensure ArduPilot Parameter, OA_DB_OUTPUT >= 2 (Send HIGH and NORMAL importance item)
      2. Obstacle icons should show up on main GUI

3. Terrain Following with Object Avoidance

The current version of ArduCopter does not allow terrain following if object avoidance is enabled.

1. Psuedo-Diagram
   

2. Walk-through

   1. Regardless whether OA is enabled, AC_WPNav_OA will always be used rather than AC_WPNav. AC_WPNav_OA::update_wpnav() will always call AC_WPNav::update_wpnav() at the end. The main problem lies within AC_WPNav_OA.
   2. AP_OAPathPlanner uses threading to process and adjust the waypoints. Sometimes, it isn't ready and will return AP_OAPathPlanner::OA_Processing.
   3. A switch case in AC_WPNav_OA handles the results. However, if the OAPathPlanner is still processing, a false boolean will disable terrain following (for that waypoint). (This might be a bug?)
   4. Since the check occurs so frequently, terrain following will be disabled for all waypoints throughout the mission.
   5. The fix is to replace the false boolean to _terrain_alt, a boolean that stores the state of terrain following, true if enabled and false otherwise.

4. Multi-Drone Simulation

To be honest, the implementation for multi-drone simulation is very inelegant and unscalable.

1. References:

   1. https://github.com/Intelligent-Quads/iq_tutorials/blob/master/docs/swarming_ardupilot.md
   2. https://github.com/Intelligent-Quads/iq_tutorials/blob/master/docs/multi_mavros_drones.md

2. Walk-through

   1. Segregrate the FDM ports of the drone model so ArduPilot (sim_vehicle.py) can correctly connect to the drone plugin.
      1. Duplicate the existing drone model
         1. iris_with_lidar -> iris_with_lidar_2
         2. iris_with_ardupilot -> iris_with_ardupilot_2
      2. Change the include model (in iris_with_lidar)
         1. <uri>model://iris_with_ardupilot</uri> -> <uri>model://iris_with_ardupilot_2</uri>
      3. Change the FDM Ports by 10 (in iris_with_ardupilot_2)
         1. <fdm_port_in>9002</fdm_port_in> -> <fdm_port_in>9012</fdm_port_in>
         2. <fdm_port_out>9003</fdm_port_out>-> <fdm_port_out>9013</fdm_port_out>
   2. Create another MAVROS instance for the new drone
      1. Duplicate the existing apm.launch file
         1. apm.launch -> apm2.launch
      2. Change certain parameters
         1. Increment fcu_url by 10
         2. Increment target_url by 1
         3. Add & Set ns to /drone2
      3. Ensure that the new drone model use the new ROS topic
         1. Duplicate the lidar 360 model
            • lidar_360 -> lidar_360_2
         2. Change the ROS topic name
            • /mavros/obstacle/send -> /drone2/mavros/obstacle/send
   3. Fix QGroundControl to allow multiple drones
      1. Navigate to the ArduPilot Repository (~/ardupilot/Tools/autotest)
         1. Duplicate the existing ArduPilot startup parameter file (in ./default_params)
            1. gazebo-iris -> gazebo-iris-2
            2. Append SYSID_THISMAV 2 to end of file
         2. Register this new parameter file (in ./pysim/vehicleinfo.py)
            • (See references)

3. Test

   1. Start the Gazebo World in ROS
      • roslaunch helium flat_multi.launch
   2. Start the ArduCopter instances
      • sim_vehicle.py -v ArduCopter -f gazebo-iris -I0 --out=tcpin:0.0.0.0:8100
      • sim_vehicle.py -v ArduCopter -f gazebo-iris-2 -I1 --out=tcpin:0.0.0.0:8200
   3. Start the MAVROS instances
      • roslaunch helium apm.launch
      • roslaunch helium apm2.launch
   4. Start QGroundControl
      • Settings -> Comm Links -> Add
        • Name: Drone 2
        • Type: TCP
        • TCP Port: 8200
      • Ok -> Drone 2 -> Connect