ncsurobotics / seawolf Goto Github PK

Catch the Ctrl+C signal so the simulator exits cleanly (currently keeps process running, must reboot or pkill to fix).

OpenCV3 is removing the old c 'cv' api entirely, and it has been deprecated for quite some time. Translate all code that is currently using the old c api to the new cv2 api.

Dependency on #20

Make GUI or other interface for better managing seawolf processes dynamically

Complete:
-- Bins

TODO:
-- update hedge
-- update gate
-- ...

Verify path mission is working

Backup all files on robot

Make the robot do an aileron roll.

Create ability to control fps of video streams in order to reduce heat problems.

Look for old SVR method to adjust fps or modify it to do so

This includes creating a way to transition between opencv and opencv2 images and recreating image transformations for use on opencv2 images.

Resolved:
-- Converting between CV1 and CV2 image formats (Kurt, Ryan & Matthew)'
-- New syntax for image transformations (Ryan & Daniel)
-- Adaptive threshold issue (Daniel)

TODO:
-- ...
-- ...

Eventually, reimplement Haar Training so the robot recognizes the LED buoys

Write code to cause settable delay when reversing direction of thruster.

This is up for debate, but mission_control should mostly be high-level logic regarding navigation. sw3 would be better off as a standalone module called 'movement' or something. I can refactor it in experimental branch if nobody has any objections.

Additionally, data.py is all IMU sensor data, it doesn't really belong in either.

Max install of wxPython is 2.8 for now. Debian Jessie repos only contain 3.0. Upgrade application manager eventually.

Create script that automatically downloads and installs everything

Amongst the 2 Cameras, IMU, and Seawolf, do we have a 5th USB port that can be used for acoustics?

Run new computer with autonomous Seawolf and all peripherals hooked up

Then mount and power new computer

Basic movement/PID testing, especially strafing.

Once the CANbus packet-dropping problem is solved, add data from leddar sensor into mission code (hopefully) better accuracy.

Rewrite gate mission to use overhead beam, not side poles in detection. This should reduce the false positives when detecting the poles.

add auto record option for missions using svr

Sensor data needs to be captured at the same time that frames are grabbed from the cameras. This was a problem in the 2011 competition with the path mission.

The solution is to send a message from vision to mission control whenever it grabs a frame. Mission control then grabs sensor values at that time.

Pitch PID needs to be updated to use variable subscriptions, like the other PIDs. Right now it doesn't even use PID_pause, or reset the integral.

SVR uses the old C api for OpenCV, which has been deprecated and is being removed from OpenCV 3. In order to remove all c api dependency, the needed parts of SVR should be recreated (likely in python).

#5

Integrate acoustics into mission code

Task

The new gate mission, gatesimp, in the newMissionBranch currently runs for a set time. Ideally the mission will wait to see the gate for a certain amount of time, if it does not see it, it will quit after that time. If it does see it, it runs until the gate is no longer seen, after that there is another short time the mission runs until it ends.

Running mission

Go to the seawolf/mission_control2 directory
do run.py
to change the missions being run modify the missions array.

Process

make new branch improveGate
make the new gateTimeState.py mission and make it work as described above.

Figure out why yawpidpy isn't working

Details:

• Regardless of whatever orientation the IMU is positioned in on the yaw plane, yawpidpy will only ever output .4 and -.4 for the port and starboard thrusters, respectively. This is totally independent of p, i, and d setpoints, as well as angle

When trying to use the application runner gui, I get the following warning

Gtk-Message: Failed to load module "canberra-gtk-module"

Everything works, but I don't think this warning should be there.

Create test that changes thruster values very quickly. Values should switch between negative and positive polarities to make sure that the thrusters don't accidentally switch too fast.

Wrote hedgeR to identify the new hedge. Currently have filters that semi accurately separate the green and red onto different frames.

Download all recorded videos off robot

With the brand-new robot, this is a good time to rewrite the missions. Also, most of them are at least a couple years old.
https://github.com/ncsurobotics/seawolf5/issues/new#

Verify Buoys Mission is working

When the value is 0 the dials show 180. at 1 or -1 the dials are okay but at 0 it shows up as 180.

Also it might be better to place the gui folder directly in the applications directory and not in applications/src. The filename also doesn't need a date stamp since github can handle the revision number.

When running missions, the cameras are always in the wrong order for the simulator.

This would be a useful change for several reasons.

• It would allow us to keep a persistant model of our surroundings and our place within them, which is a technique used by Cornell, Sonia, and UF.
• All of the sensor data from cameras, acoustics, (leddar, DVL, sonar?) can be combined into one unified model of the environment and reduce the amount of crap that mission control has to deal with.
• It would make implementing new sensors much easier, since they can just contribute more information into that model instead of trying to handle it separately