Welcome back to another action-packed, thrilling pool test! As has become tradition for the past few months, we only brought Oogway to the pool test, with Crush still in quite a discombobulated state. As such, we had an ambitious list of features to test.
Prequalify
The first thing we tried today (erm, actually this was almost 3 weeks ago, oops) was prequaling; we ran into a lot of errors, mostly stemming from the fact that we didn’t get any Lane Marker detections. However, we also had random issues during runs during initial submerge, and move_to_pose commands timing out. Further debugging (before we abandoned it to try other tasks) revealed that we weren’t getting any HSV detections at all, which might be an issue with refactoring. Before giving up, we tried running a dead-reckoning version of prequal, which failed miserably. I guess we do need CV for something after all.
Sonar
Sonar was one of the brighter sides of this pool test. We tested the angle of approach pipeline by reading the pool walls at different angles, and got pretty accurate results without bugs. We seem to be approaching the end of Sonar processing development, and it looks like this can be a part of our task planning pretty soon!
Yaw to CV Object
We finally tried running the new Yaw to CV object, but ran into bugs with 5 minutes remaining in the pool test. 🙁
Acoustics
We were unable to find an opportune moment to test acoustics.
DVL Disconnect
We tried to see if the DVL was disconnecting throughout the pool test, but we weren’t able to attribute the “freezing up” of DVL readings to foxglove or to the DVL itself. To solve this for future pool testing, we decided to create a simple ros2 node that logged DVL readings to a file directly.
So all in all, a pretty eventful (and bug-filled) pool test. 6/10.