2024-01-11 Autonomy : Directors sync
Projects (W24 focused)
W24 status table, using 2023-2024 Project List and discussion with @Mihir Gupta @Amy Hu
Detailed notes + fwd progress below table
Project | Timeline/progress | EOT goal |
---|---|---|
Integrated Monitoring And Command Station (IMACS) 2.0 | ongoing probably not comp critical | continue to expand GUI |
Airside System | geolocation in flight test
| integration tasks - especially geolocation |
Pathing | Task 2 program done - need to test this, will do simulator first Working on task 1 | Both task 1 and 2 working and tested |
ML model | have datasets, collected over fall may need to flight test more to collect more data for the learning model | cleanup for comp? - make a timeline for this mid-to-late feb |
LTE Communication | profiling image compressions - seeking best version for LTE comms purchasing Pi for testing | see timeline below |
IMACS
Current:
gui updates from vals over mavlink
Goals:
Update gui ++ put more members on
Notes:
Good proj to have, not enough “comp” tasks to give to everyone. IMACS good for getting new members to contribute to warg.
Airside system
Current:
Goal for this term:
Flight test geolocation + estimation.
Unsure if full airside system will be working before competition.
Need to start working on integration tasks with geolocation.
can easily be thrown off by errors in system.
(probably req. lots of ft)
next ft: geolocation. Aiming to work w/team members to integrate.
Want to do ground test first, possible to ground test [anticiptaing february] + simulation.
once ground test working, can continue on to flight tests.
Geolocation may not be ready for this comp, but important for future competitions.
FT timeline:
expecting february
Pathing
Need to test in sim, and then flight test.
Task #2 working, working on dev. Task #1.
Not much progress on task #1. Could be tested on sim by april.
FT timeline:
FT task #2 starting February
FT Task #1 starting March
Continuous flight testing in diff. conds over remain. ft’s
ML Model
Bunch of rounds of data collection done
Unclean data set ready
Now needs to assign tasks out to clean up the data and annotating it on roboflow so model training can begin.
Progress good, begin training model soon. data cleaning not expected to take too long.
Once dataset cleaned up, need to see how many images needed.
May want to ft more to collect more data for model.
FT timeline:
Can ground test model by moving camera
Plan is to integ w/airside system at mid-end february
possible to use last years model until then
Depends on LTE being ready.
Continuous FT for image collection
LTE
Able to stream drone telemetry to groundside
Focus for this term: transmitting images from camera,
have started profiling different image compression formats. Trying to determine what is ideal
Purchase request for another pi to start testing what the code does on the rpi.
Once compression determined, create repository that runs on the rpi.
Current plan (up for debate / changes)
Get messages from FC when mission is done
Done through dronekit - PR up, code is done and working on testing it with mission planner simulation. Should be done soon.
When done, start transmitting images over to groundside
Have proven possible, want to pick optimal compression.
Groundside will run & upload landing pad location to drone (telling it where to go)
Have determined possible to determine when mission is done (will send command to loiter over certain lat/lon)
Jetson → Pi → FC
FT Timeline:
verify LTE works on ground (date: mid-feb)
LTE ↔︎ Airside ↔︎ ML model ground (date: start of march)
Fully integrated airside test (date: )
LTE + airside + ML model → complete autonomous landing task. worth 5 pt.
Task 1 code pathing → worth 15 pt. (approx)
emphasize task 1 code to validate this for competition. more important for competition than the first 3 items. does this change timeline?
task 2 highest priority for flight testing
task 1 code priority over auto landing
Megans Questions
AEAC Sync
How do you like the format of AEAC sync? Is the content effective? useful?
AEAC syncs go through a spectrum, mech → ee → efs → auto
most discussions that each subteams are having is good and productive
Wanted to suggest instead of going mech → ee → efs, could go the other way around
trying to get auto members to show up, but mech/ee/efs isn’t as important for them and auto members get bored.
can be a drag if you’re just a team member
organize by projects instead?
better - > each subteam more involved with each section of the meeting
also helps PMs have more context related to integration
Leads meeting
Last term leads meeting was mostly admin/policy updates and became quick. Was that a reasonable format for leads meetings? Was that delivering what you thought leads meetings should be delivering? Is there anything you would’ve liked to change?
Leads meetings were good, leads tend to be drawn into other meetings, keeping leads meeting short and concise is very helpful.
Whenever there is discussion @ leads meeting it’s relevant and productive, really like the current leads meeting structure.
Useful to hear about admin info.
Since a lot of leads / involved members go to AEAC sync, would you consider folding / mergeing AEAC & leads meetings together?
No - aeac sync is important and can be very draining (heavily involved in a lot of technical discussions and problem solving). Booking another meeting after that would not be very productive
Do you think leads needs to be in person?
Having the option is good, i.e. if people are working in the bay / campus and want to come in person it is good (for being social / cultural)
Depending on locations, schedules, should continue to hold it hybrid.
Director support
For people not autonomy, autonomy system feels like a black box. Want to have people understand the autonomy system more, and not let it be a black box.
If anyone wants to know how autonomy system works (and not buzzword), more than happy to explain and make it understandable / approachable.
Try and run socials!
New leads for auto?
bringup 2 leads, mentor & onboard them.
become subteam advisors (want to try different roles/subteams).