vortexntnu / vortex-auv Goto Github PK

The control mode is updated through the /propulsion_command topic now. To bypass this topic we want to create a rosservice to change the control mode

RuntimeError: Multiple packages found with the same name "imu_complementary_filter":

• manta_ros/manta-auv/navigation/imu_complementary_filter
• manta_ros/stim300/imu_tools/imu_complementary_filter

Multiple packages found with the same name "rviz_imu_plugin":

• manta_ros/manta-auv/navigation/rviz_imu_plugin
• manta_ros/stim300/imu_tools/rviz_imu_plugin

The offending packages in vortexntnu/manta-auv should be removed or at least renamed

Ambjørn has the Sonar driver. The sonar driver needs to be in a private repo. It can not be public, as the company we got it from requires it to be private.

Can we reduce the buoyancy to match real world behavior @Sollimann

To stick to our want of abstracting past the manta auv, the topics that are currently named /manta/... should be changed to something more generic like /auv/...

try google it: "change color of mesh in .xacro files" or something similar.

We need to be able to provide the setpoints necessary to go through the gate, around the pole and back through the gate.

These files are redundant or inaccurate

Add further files in the comments.

1. MANTA-AUV/control/vortex_allocator
   Reason: using yaml-files written for Terrapin and ROV-Manta

2. MANTA-AUV/control/vortex_controler/test/controller_test.test
   Reason: using yaml-files for maelstrom

3. MANTA-AUV/install-prereq.sh
   Reason: using kinetic and not melodic

Github Actions can be used to set up CI. There is an action on the Actions marketplace specifically for ROS environment setup, so it might be doable without too much work

We are testing two allocators that returns forces to the thrusters but they are returning very different thrusts. I suspect that this might cause the problems with the simulator when trying to jaw and surge at the same time.

When using a joy stick and providing max heave the output thrust from allocator is uuv_allocator = 34.09 while our_old_allocator = 7.20. Similar behavior is observed in the other DOF.

While doing this, the manta/thruster_manager/input is wanted total thrust = 28.8 this suggests that that the uuv_allocator is wrong since 28.8/4 = 7.2 unless the simulator does something fancy to neutralize the offset after after the allocator.

@Sollimann Why is max_thrust = 2040 in manta-auv/uuv_simulator/manta_control/manta_thruster_manager/config/thruster_manager.yaml

Unable to load executive_smach_visualization as it is not a submodule.

The setups for Manta and Gladlaks should be added as separate pages to the wikis user manual.

Most of the required steps for Manta are already described in the wiki on teams, and can be used as a starting point.

The AUV is supposed to be able to launch torpedoes, so this needs to be implemented in software. This can only be finalized once the hardware configuration for the launch mechanism has been decided.

The camera centering brannch contains a lot of new code, but it will likley not be needed for the new AUV. There are also files that should be removed from the branch, eg. compiled python files.

Task: either delete branch or clean it up and commit it

Create an overview of existing path-following controllers. List pros, cons and requirements for each.

It would be helpful to have a readme in all top level folders that explains what the folders should contain and what its submodules are.

Calculate distance to gate and pole using height of boundary boxes in computer vision node and publish it on /camera_object_info topic.

Alter /CameraObjectInfo msg type to include distanses.
Implement real height and ratio as global parameters.

The current state machine is very bare-bones, and needs to be expanded with the current RoboSub 2020 tasks. The current structure of the state machine can also be updated if there are betters ways of structuring it.

Once the RoboSub 2020 tasks are implemented and this issue is closed before the 2021 tasks are revealed, there needs to be a new issue addressing updating the state machine to fit the new tasks.

If the RoboSub 2021 tasks are revealed before this issue is closed, the state machine will have to be updated to the new tasks before closing the issue.

The thruster models used in uuv_simulator is larger than the bluerobotics T200 thrusters.

Proposed fix: Find T200 CAD files online and convert to .dae file and change the manta model

State-machine:
skru av og på noder
Som mappe i manta_auv: state_machine

Oppgaver

Initialize
Aktiver depth hold
Aktiver heading hold
Gå til 1m dybde

Finn gate - For ROBOSUB
Søke i heading - Aktiver search pattern

posisjoner
Sender ut goal node
Camera centering (heading + depth hold + x + y)
depth hold

gå gjennom gate
generer punkt forbi gate
Gå til punkt

Søke pole
Heading søk

Gå rundt pole
Basic alternativ: Camera centering
Mer avansert: Generere punkter og følge

returner til gate, gå gjennom gate
Basic alternativ: Camera centering
Mer avansert: Generere punkter og følge
Feede gamle punkter tilbake

surface
Gå til overflaten

stopp
shut down

A report for the previous work on acoustics needs to be written before this year's work on the acoustics.

The report should contain:

• What has been done in the previous year
• What worked or went well?
• What did not work or went badly?
• Any other valuable experiences from the previous work
• What needs to be done this year

The parameters r, r_d and r_d_dot are currently undocumented in the LOS and autopilot nodes, since no good explanation for them has been found.

These parameters need to be documented.

The directory structure can be confusing, and should be restructured into a more intuitive format. A new format should be decided on, and a first proposition is provided below:

Motion: thruster control and path following
Navigation: Localization, mapping and path planning
Mission: State machine and states
Arm: Controller for arm
Object Detection: Detection of objects visually and with acoustics

The old controller need to operate on setpoints provided

Not able to kill camera_centering node with ^C

Before this task can be handled, make sure we have gotten the driver from Ambjørn and make sure the issue is marked as closed.

Write a roswrapper for the Sonar driver and modify the code in Vortex-AUV to reflect the change.

Sensor synchronization is common in systems which depend on multiple sensors, and we should examine if it is something we want to implement

We need a rosservice that saves the local position (x_tasknr, y_tasknr, z_tasknr) of each task. This allows us to generate waypoints for our path planning.

There are plenty of old branches in this repo and the simulator, which should be removed

ORBSLAMv3 is a state of the art SLAM system the Vortex is planning use.

Tasks:

• Install ORBSLAM3 locally and test it with a public dataset
• Integrate ORBSLAM3 into Vortex-AUV

The object detection, stereo camera, and mapping system requires testing data from an underwater camera rig to be tested properly. This footage needs to be gathered.

We should have all tunable parameters in a yaml file for ease of life purposes. If possible it could be nice with dynamic configuration. Bør og ha parametre for max/min til thruster.

Goal: Simplify codebase and improve maintainability

All cases where there are imports accross packages should be replaced with ros interfaces, such as topics.

Known cases so far: guidance/los backstepping imports controller/vortex_controller

Certain packages have files in them used for testing the package separate from the rest of the system, and this has led to some clutter. Issue #70 refers to both of these cases; having a /test folder in every package that needs one is therefore good for uniformity and cleanliness in the codebase.

Simulated pressure sensor is not implemented properly and therefore depth_hold is not working.

Descriptions of Manta are currently published as parameters that are loaded from yaml files during launch. The yaml files are located in control/vortex/config.

URDF is the standard for how descriptions of robots should be handled. Our simulator is already using the format, and we should consider adopting it in our main repo as well. This change has been recommended by Kristoffer Solberg.

To use OpenCV etc. with the simulator we need to extract the simulated front camera view.

Already done in the old simulator so probably not that difficult now.

@Sollimann have you looked at this already?

Now we need to run 5 different commands to launch the complete simulator. (including rqt for camera view)

It should be possible to reduce this to 2 commands. One to launch simulator and one to launch controller.

Delete unnecessary files and restructure.

Easier to copy/paste without the "$" signs

Find existing solutions on depth estimation of detected objects from camera.
Consider if point cloud should be used for depth estimation.

One of the hydrophones diverged from the other two last year, so we should check if it is broken. This is important, as one hydrphone costs ~10k NOK

The setpoints for controller gets updated automatically when using a joystick and the topic /propulsion_command. We want to be able to set these setpoints with a rosservice.

I recommend starting with setpoint for depth hold mode, then extrapolate to heading and x,y

Find existing packages for mapping and dokument the following

• Requirements
• Limitations
• Pros and cons

Feel free to update the REAMDE.MD whenever you have some useful information or instructions about a particular package.

There is a lot of unmerged work on the path-planning branch, but it need to be cleaned up before it can be merged

Tasks

• Remove .vscode files
• Remove compiled python files
• Fix merge conflict

The README is not updated for melodic, and needs to be linted for unnecessary/wrong steps.

ORBSLAMS performance can be improved by adding DVL and depth measurements. The existing version uses only IMU.

This task is probably very challenging and requires good knowledge of c++.