This repository contains packages for simulating the Tethys-class Long-Range AUV (LRAUV) from the Monterey Bay Aquarium Research Institute (MBARI).
Disclaimer: This repository is in active development. Stability is not guaranteed.
Source files, models, and plugins relevant to a general audience are upstreamed on an irregular basis to Ignition libraries, the top-level library being ign-gazebo. Upstreamed files may eventually be removed from this repository.
Standalone, this repository contains the environment and plugins necessary to simulate an underwater vehicle in Ignition Gazebo. Integrated with the real-world LRAUV controller code, the binaries of which are provided to the public on MBARI's DockerHub (see below), the simulated robot can be controlled using the same code executed on the real robot. This enables the validation of scientific missions for oceanography research.
Optionally, you may choose to build this repository using Docker, for convenience. Make sure you have a recent version of Docker and nvidia-docker installed. Next to get started simply run the following command.
docker/build_and_run_docker.sh
To join in a separate terminal, remember to source Ignition and the workspace:
docker/join.sh mbari_lrauv
. /home/ign_ws/install/setup.bash
. /home/colcon_ws/install/setup.bash
To run the code in this repository natively without Docker, make sure you have Ignition Garden and colcon, on Ubuntu Focal or higher.
Install dependencies
sudo apt-get install libpcl-dev
Clone this repository then run
colcon build
Developers may want to build tests. Note that this would take longer:
colcon build --cmake-args "-DBUILD_TESTING=ON"
You can pass
--cmake-args ' -DENABLE_PROFILER=1'
to use the profiler. See more on this tutorial
This package comes with an empty example world. To run this example world simply source the colcon workspace and run:
. install/setup.bash
ign gazebo buoyant_tethys.sdf
Send example commands to move some joints:
LRAUV_example_controller
Keyboard teleop:
LRAUV_keyboard_teleop
Tip: Type
LRAUV_
and press tab for autocomplete to show more example examples.
MBARI's code base lives in a separate, private repository. For people with access, here are instructions on setting it up from source and compiling.
Alternatively, you can use the public Docker image (see below).
The integration assumes that this repository is cloned as a sibling of
the lrauv-application
repository, i.e.:
<workspace>
|-- lrauv
â””-- lrauv-application
For quick reference, compilation boils down to running this on the right branch:
make ignition
A public Docker image is available for people without access to the MBARI codebase. MBARI's image on DockerHub contains Ignition, MBARI's LRAUV code base, and this repository. All the code has been compiled.
docker pull mbari/lrauv-ignition-sim
Note: To update that image, see
instructions
in MBARI's private lrauv-application
repository.
Once inside a container, source the colcon workspaces:
. ~/ign_ws/install/setup.bash
. ~/lrauv_ws/install/setup.bash
This needs to be done for each terminal.
For ease of development, the following world is set up to run at a real time factor (RTF) of 0 (as fast as possible) and a step size of 0.02 seconds. That is significantly faster than the default Ignition setting of RTF 1 and step size 0.001 seconds, which will give real time performance and roughly the nominal vehicle speed of 1 m/s.
The RTF and step size can be changed at run time via the GUI by going to the
Inspector panel and then Physics Group.
Alternatively, they can be changed prior to compilation in the world SDF under
<physics><max_step_size>
and <physics><real_time_factor>
.
Launch the Ignition simulation:
ign gazebo buoyant_tethys.sdf
For verbose debug output, add --verbose 4
.
Unpause Ignition by clicking on the triangle play button in the lower-left corner of the GUI, or by pressing the space bar.
The MBARI LRAUV Main Vehicle Application (MVA) contains everything needed to control and operate the vehicle in the real world and in simulation.
Time synchronization has been done between the MBARI application and the Ignition simulation, such that in most cases, the controller and the simulation should be running in synchronization. There may be corner cases that still need to be resolved.
This section assumes you have either compiled the target from source or are using the MBARI public Docker image, which has everything pre-compiled. The paths assume you are using the MBARI public Docker image.
Run the LRAUV Main Vehicle Application:
cd ~/lrauv_ws/src/lrauv-application
bin/LRAUV
This will bring you to a command prompt.
At the LRAUV command prompt (you only need to do this once):
>configset micromodem.loadatstartup 0 bool persist
>restart app
This sets the micromodem to not load at startup.
persist
means you only need to do this once.
It will pause for a bit, you might not be able to type right away.
On the vehicle, an app is always being run.
If no missions are specified, then it is running the default.
On the real vehicle, the default mission is GoToSurface
.
(NOTE: we have removed the default app in the MBARI public image until
this issue is resolved.
Currently, nothing is being run by default. Skip this check.)
Verify that it is running the default GoToSurface
app:
>show stack
2021-03-03T18:24:46.699Z,1614795886.699 [Default](IMPORTANT): Priority 0: Default:B.GoToSurface
Control commands can be issued to overwrite mission controls. For example, the rudder can be held at a constant angle like so:
>maintain control horizontalcontrol.rudderangleaction -15 degree
This overwrites the controller and maintains the rudder at -15 degrees (-0.261799 radians), which is the joint limit.
Unit conversions are automatically done in the MBARI code. Alternatively, you can specify in radians.
>maintain control horizontalcontrol.rudderangleaction -0.2 radian
A thruster command can then be issued to move the vehicle in a circle:
>maintain control SpeedControl.propOmegaAction 300 rpm
Currently, this is the tested and preferred method of control.
A sample list of command variables:
Config/Control-->HorizontalControl.loadAtStartup=1 bool
Config/Control-->HorizontalControl.kdHeading=0.049999 s
Config/Control-->HorizontalControl.kiHeading=0.001000 1/s
Config/Control-->HorizontalControl.kiwpHeading=0.000500 rad/s/m
Config/Control-->HorizontalControl.kpHeading=0.400000 n/a
Config/Control-->HorizontalControl.kwpHeading=0.049999 rad/m
Config/Control-->HorizontalControl.maxHdgAccel=7.499876 arcdeg/s2
Config/Control-->HorizontalControl.maxHdgInt=0.087266 rad
Config/Control-->HorizontalControl.maxHdgRate=11.999932 arcdeg/s
Config/Control-->HorizontalControl.maxKxte=45.000001 arcdeg
Config/Control-->HorizontalControl.rudDeadband=0.500000 arcdeg
Config/Control-->HorizontalControl.rudLimit=15.000000 arcdeg
VerticalControl-->VerticalControl.buoyancyAction=944.986938 cc
VerticalControl-->VerticalControl.depthIntegralInternal=nan rad
VerticalControl-->VerticalControl.depth2buoyIntInternal=nan cc
VerticalControl-->VerticalControl.massIntegralInternal=nan m
VerticalControl-->VerticalControl.elevatorIntegralInternal=nan rad
HorizontalControl-->HorizontalControl.rudderAngleAction=0.000000 rad
SpeedControl-->SpeedControl.propOmegaAction=0.000000 rad/s
The following are "unit test" missions that test one or two actuators at a time.
Run one at a time, in separate runs of Ignition and the Main Vehicle Application
(bin/LRAUV
):
run RegressionTests/IgnitionTests/testDepthVBS.xml
run RegressionTests/IgnitionTests/testPitchMass.xml
run RegressionTests/IgnitionTests/testPitchAndDepthMassVBS.xml
run RegressionTests/IgnitionTests/testYoYoCircle.xml
Some example behaviors are documented here.
Some parameters can be adjusted - see the mission XML file.
For example, to change the commanded depth in the testDepthVBS.xml
mission:
load RegressionTests/IgnitionTests/testDepthVBS.xml
set buoy_test_vbs.DepthCmd 20 meter
run
To stop a mission, run
stop
You can automate typing into the command prompt by issuing -x
.
For example, this will run the yoyo mission and terminate after the mission ends:
bin/LRAUV -x "run RegressionTests/IgnitionTests/testYoYoCircle.xml quitAtEnd"
The original simulation (SimDaemon
) is the baseline comparison for the
Ignition simulation.
For developers, it helps to troubleshoot the Ignition simulation by comparing its values to the original MBARI simulation, which is a pure command-line interface.
Do not run both Ignition and SimDaemon at the same time. Choose only one.
cd ~/lrauv_ws/src/lrauv-application
In the MBARI code base, open Config/sim/Simulator.cfg
, change these lines to
look like this:
ExternalSim.loadAtStartup = 1 bool;
ExternalSimIgnition.loadAtStartup = 0 bool;
This enables the original ExternalSim and disables the interface with Ignition.
Run the original command-line simulation:
bin/SimDaemon
The SimDaemon runs in the background by default.
Then run the Main Vehicle Application as usual:
bin/LRAUV
Speed up 100 times for a bit to finish loading, before returning to normal speed. This allows the commands to finish loading, before you overwrite them with control commands. Otherwise the preloaded commands can kick in after you issue control commands and make the vehicle go to unexpected places
>quick on
>quick off
Alternatively, if you have access to the config files, set SBIT.loadAtStartup to 0 bool in Config/BIT.cfg. This might already be set for you in the Docker image on MBARI DockerHub.
Load the circle mission, which will perform two circles:
load Engineering/circle_test.xml
Set some parameters as desired:
set circle_test.Depth01 10 meter;set circle_test.Depth02 15 meter;set circle_test.RudderAngle01 15 degree;set circle_test.RudderAngle02 10 degree;set circle_test.WaitDuration 10 minute
run;quick off
You can check variables like depth:
report touch depth
quick on
To clear the report and go back to normal speed:
report clear
quick off
To stop the mission and terminate:
stop
quit
A simple dockerfile and tmux config exists that makes launching and debugging the different components of the project a lot easier. To use it simply run
$ docker/debug_integration.sh
This will build a new container with the source code and launch a tmux session. The tmux session has 2 windows: 0:simulation and 1:logging. In the simulation window you will see the top pane runs the ignition simulation while the bottom pane runs the actual bin/LRAUV
controller. The logging pane on the other hand will automatically convert the sim slate and write it to the results directory on your computer one layer above the directory to where you checked out.
This contains some most-often used commands for quick reference:
Show general help or for a specific command:
>?
>help report
Show mission currently being run
>show stack
Speed up 100 times faster than real time:
>quick on
# To go back to normal speed
>quick off
To report the value continuously on variable touch:
>report touch <componentName>.<variableName>
# To stop reporting
>report clear
To get the current value of a variable:
>get <componentName>.<variableName>
A sample list of variables in the ExternalSim
component:
ExternalSim.latitudeSim=36.803400 arcdeg
ExternalSim.longitudeSim=-121.822200 arcdeg
ExternalSim.eastingSim=605067.311028 m
ExternalSim.northingSim=4073710.248871 m
ExternalSim.utmZoneSim=10 enum
ExternalSim.propThrustSim=-0.000000 N
ExternalSim.propTorqueSim=-0.000000 N-m
ExternalSim.netBuoySim=0.000000 N
ExternalSim.forceXSim=0.000000 N
ExternalSim.forceYSim=0.000000 N
ExternalSim.forceZSim=0.000000 N
ExternalSim.posXSim=0.000000 m
ExternalSim.posYSim=0.000000 m
ExternalSim.posZSim=0.000000 m
ExternalSim.rollSim=0.000000 rad
ExternalSim.pitchSim=0.000000 rad
ExternalSim.headingSim=0.000000 rad
ExternalSim.posXDotSim=0.000000 m
ExternalSim.posYDotSim=0.000000 m
ExternalSim.posZDotSim=0.000000 m
ExternalSim.rateUSim=0.000000 m/s
ExternalSim.rateVSim=0.000000 m/s
ExternalSim.rateWSim=0.000000 m/s
ExternalSim.ratePSim=0.000000 m/s
ExternalSim.rateQSim=0.000000 m/s
ExternalSim.rateRSim=0.000000 m/s
ExternalSim.homingSensorRangeSim=27.335945 m
ExternalSim.homingSensorAzimSim=-1.531450 rad
ExternalSim.homingSensorElevSim=1.073800 rad
To stop a mission:
>stop
To terminate:
>quit
Each instance of bin/LRAUV
is tied to a single vehicle. In order to work
with multiple vehicles, multiple instances of bin/LRAUV
must be spun up.
The first vehicle spun up will be placed at the origin of the world, and the latitude / longitude of the world's origin will be set to coincide with it. Subsequent vehicles will be spawned at positions relative to the initial one, according to their latitude / longitude.
Information about the vehicle is hardcoded on the lrauv-application
code,
within the Config
folder. Here's a recommended setup assuming that you have
lrauv-application
cloned under ~/lrauv_ws/src
:
-
Copy the
lrauv-application
folder for each robot to be spawned:cp -r ~/lrauv_ws/src/lrauv-application ~/lrauv_ws/src/lrauv-application-2 cp -r ~/lrauv_ws/src/lrauv-application ~/lrauv_ws/src/lrauv-application-3 ...
-
Edit the vehicle name (in
Config/vehicle.cfg
) and initial location (inConfig/workSite.cfg
) for each instance. For example:--- lrauv-application/Config/vehicle.cfg 2021-09-27 16:17:09.816305451 -0700 +++ lrauv-application-2/Config/vehicle.cfg 2021-09-29 14:53:57.480185748 -0700 @@ -10,7 +10,7 @@ //////////////////////////////////////////////////////////////////// // Vehicle.name = "Tethys"; - Vehicle.name = "tethys"; // Use name to match Ignition default SDF + Vehicle.name = "daphne"; // Use name to match Ignition default SDF Vehicle.id = 0 enum; Vehicle.kmlColor = "ff0055ff"; // 4 hex bytes indicating alpha, blue, green, and red // In this case, orange. --- lrauv-application/Config/workSite.cfg 2021-09-27 14:16:43.622409403 -0700 +++ lrauv-application-2/Config/workSite.cfg 2021-09-29 14:53:06.887476472 -0700 @@ -14,8 +14,8 @@ // initLat = 36.806966 arcdeg; // Initial latitude // initLon = -121.824326 arcdeg; // Initial longitude // initial position same as for regression tests - initLat = 36.8034 arcdeg; // Initial latitude - initLon = -121.8222 arcdeg; // Initial longitude + initLat = 36.8033 arcdeg; // Initial latitude + initLon = -121.8223 arcdeg; // Initial longitude startupScript = "Missions/Startup.xml"; // Mission to run on power-up defaultScript = "Missions/Default.xml"; // Mission to run when no other mission is running.
To run simulation, use the empty environment, and start vehicles in order, for example:
ign gazebo empty_environment.sdf -v 4
~/lrauv_ws/src/lrauv-application/bin/LRAUV
~/lrauv_ws/src/lrauv-application-2/bin/LRAUV
- Start more
bin/LRAUV
as needed
Some worlds support levels.
Levels are turned off by default, which means that all heightmap tiles will
be loaded at all times. When levels are enabled, only the tiles containing
vehicles (performers) spawned with WorldCommPlugin
with an
lrauv_init
message will be loaded.
For example, loading without levels:
ign gazebo -v 4 portuguese_ledge.sdf
And with levels:
ign gazebo -v 4 portuguese_ledge.sdf --levels
After issuing control commands, for example, rudder and thrust, if you then notice that the vehicle gets some commands by itself, such as a non-zero elevator angle, this is because a preloaded mission is being loaded, and you need to wait to issue the control commands after it is done loading. Make sure to use
quick on
to let the system finish loading, before issuing control commands.
On the MBARI Main Vehicle Application side, all values during the run are stored to disk. They can be retrieved after the run and plotted for debugging purposes.
See lrauv_ignition_plugins/plots/README.md
for instructions to unserialize and scripts for plotting.
Science data can be read from a csv file with the following recognized field names in the first line of the file:
elapsed_time_second
latitude_degree
longitude_degree
depth_meter
sea_water_temperature_degC
sea_water_salinity_psu
mass_concentration_of_chlorophyll_in_sea_water_ugram_per_liter
eastward_sea_water_velocity_meter_per_sec
northward_sea_water_velocity_meter_per_sec