Versions Compared

Old Version 24

changes.mady.by.user Vibhinn Gautam

Saved on

compared with

New Version Current

changes.mady.by.user Ashish Agrahari

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Table of Contents
stylenone

Planning

Scope

  • Houston

    • Testing new camera

      • We have a non-global shutter RPi Camera we need to test on the drone

      • Autonomy should try geolocation as well, need to get script ready

      • Should acquire data of just images (with nothing overlayed) so we can test algorithms

      • Note: Geolocation with the CV camera will be tested FIRST as it requires more setup. Then the CV camera will be swapped out for the IR camera, and the IR camera will be used to capture images.

    • Tracking Antenna ~ Moved to next flight test

      • Must be tested and verified before 2024-11-06 to include in this flight test Omer Sajid

      • Pre Flight Test validations

        • Test drone selected - Houston chosen

        • Determine and setup link OTA link on Houston, RFD or ELRS

        • E2E Mavlink connection established (drone → mavproxy → mission planner + tracking antenna)

        • Basic tracking works (tracking antenna follows drone being walked in a 180° arc around it)

      • If ready, can be tested alongside another Flight Card to decrease necessary flights

    • Testing object detection?

Flight Cards

Flight Card 1

Autonomy Geolocation

Pilot: Evan Janakievski

GSO: Sam Zhang Yuchen Lin

Note

What is being tested? To be filled out by flight card requesters Ashish Agrahari Maxwell Lou

  • Geolocation + Cluster estimation accuracy

    • Have more precise measurements of the location of landing pad relative to the drone take-off position.

    • Have a marker for starting location so we can put drone back there every time

      • Preferably switch on drone at starting location, or ensure that starting location is logged

    • Test different heights (ie 3-5m, 10m, 15m, 20m), each should be its own independent flight

    • Test different speeds? (optional, just go approximately same speed/slow if not testing this)

  • Local logging should work

Info

Hardware to be installed (change status to success when installed or plan for installing at flight test is done)

  • CV camera

  • RPi

THINGS TO BRING

  • Measuring tape

  • Marker/tape/box or something to mark drone take-off position

Geolocation + Cluster Estimation Accuracy Procedure (HOUSTON)

Procedure

What marks a success?

Knockoff criteria

Goals/Objectives

Knockoff criteria

  1. Setup drone: (can be done at or before going to testing location?)

    1. Switch on drone, ensure the GOOD cv camera is connected

    2. (at time of writing, it is unclear if rc.local autostarting will work). SSH into pi, start airside

    3. Read live log output, compare flight controller log home location coordinates to phone’s gps

    4. Point drone’s cv camera at landing pad (keep a couple meters distance), check if geolocation is logging and data merge logs timestamps of detections and odometry

  • A good outcome is if:

    • flight controller log has correct home location and altitude (compare with a phone’s GPS)

    • drone is able to detect landing pad and geolocation logs the coordinates (also ensure timestamps are logged in data merge logs)

< list of reasons why you would want to stop the test >

  • Geolocation Logs are not working

  • If data merge or flight controller logs not working, leave decision up to the tester present?

  1. Either

    1. Switch on drone at takeoff location (Start airside system when drone at takeoff location)

    2. Get coordinates of drones takeoff location

  2. Get distance from home location to landing pad (measure it)

    1. OR takeoff on landing pad

  • Good outcome: Flight controller log has the home coordinates

  • Able to manually measure relative distance from takeoff location to landing pad

  1. Flying back and forth over the landing pad at different heights (3-5m, 10m, 15m, 20m)

    1. Each time a height changes, airside system should be restarted (each test should be an independent flight)

  • Good outcome: images show bounding boxes over landing pads, geolocation logs are outputting detection coordinates

Flight Card 2

IR Camera Test

Pilot: Evan Janakievski

GSO: Sam Zhang Yuchen Lin

Note

What is being tested? To be filled out by flight card requesters Maxwell Lou Vibhinn Gautam

  • Collect test/training data using IR camera

  • Fly at different heights (ie 3-5m, 10m, 15m, 20m)

  • Fly at different speeds (ie for each height, do 3 slow passes and 3 fast passes)

    • These don’t need to be separate flights, can be done all in one go

Info

Hardware to be installed (change status to success when installed or plan for installing at flight test is done)

  • IR Camera

  • RPi

Geolocation + Cluster Estimation Accuracy Procedure IR Camera Test (HOUSTON)

ProcedureWhat marks a success?

Goals/Objectives

Knockoff criteria

  1. Setup (prepare drone)

  • A good outcome is if rc.local succesfully runs and images are auto captured and stored in a designated folder

< list of reasons why you would want to stop the test >

  1. Hover at 3-5.

  • Best outcome: Image output is desirable (high quality and emitter light is clear visible with no discernible issues in image quality) and are successfully stored in a logging folder.

  • Acceptable outcome: Images are successfully stored in a logging folder (we will see what why image quality is poor).

due to vibration on drone, images are very blurry.

Weather: cloudy

  1. Hover at 10m

Same as above

  1. Hover at 15m

Same as above

  1. Hover at 20m

Same as above

Flight Card 3

...

Obstacle Detection - Testing GuidedMode.

Pilot: Evan Janakievski

GSO: Sam Zhang Yuchen Lin

Note

What is being tested? To be filled out by flight card requesters Andrew Shum

  • Upload a default mission, and then use guided mode to interrupt that mission (ie after some time)

  • After Guided Mode instructs the drone to move somewhere else, it should switch back to AUTO Mode and continue its original mission.

Info

Hardware to be installed (change status to success when installed or plan for installing at flight test is done)

  • Lidar

  • RPi

Geolocation + Cluster Estimation Accuracy Procedure Obstacle Detection/Guided Mode (HOUSTON)

Procedure

What marks a success?

Knockoff criteria

Goals / Objectives

Knockoff criteria

  1. Setup drone:

  2. a) Ensure the guided-mode script has been uploaded to the drone

  3. b) Switch on drone

  4. c) Run the script (starts the mission, where drone navigates to waypoints in a square, switches from AUTO → GUIDED → AUTO, completes mission and returns to landing)

  • Good outcome if drone completes the auto-guided-auto mode switches seamlessly, drone continues its auto mission after diversion, and returns to land.

IMP: pls see photo before reading the description below!

  • The mission is setup as follows. connection initialized, armed, takes off 10m altitude, flies 10m NW to the first vertex of the square, flies 10m E to the second vertex. When the drone is about to arrive at second vertex, it switches from AUTO to GUIDED to simulate a diversion and navigates to a waypoint 7.5 NE of current pos. After reaching this intermediate wp, drone switches to AUTO, continues to 3rd waypoint. Just before it reaches the third waypoint there is another auto-guided-auto switch, abrupt change in velocity component direction. Then we carry on naturally to 4th wp. halfway between, we auto-guided-auto (through ned_waypoint this time) to another intermediate wp (see photo), return to waypoint originally navigating too, and initiate landing procedure.

  • 10m S, 4th waypoint 10m W, signifying the completion of the auto-mission. The drone then returns to landing position and lands gracefully.

< list of reasons why you would want to stop the test >

Pilot intervention is not necessary unless safety issues arise, or there are environmental disruptions, as the auto → guided → auto switch is handled by the script.

if at any point the pilot needs to take control: 1) kill the script, this forcefully changes drone mode to loiter (tested on MP), and then aborts the current mission. 2) Pilot has manual control over the drone

NOTE: on mp, simulating a mission-abort, ie. killing the script changes mode to loiter almost instantly, but drone altitude plummets to zero. (discussed with Andrew and it should just be a MissionPlanner default issue). However, just in case, pilot should be ready to control vertical throttle when aborting mission (when killing the script)

image-20241116-020108.pngImage Added

Flight Card 4 ~ Moved to next flight test

Tracking Antenna Test

Pilot: Evan Janakievski

GSO: Sam Zhang

Note

What is being tested? To be filled out by flight card requesters Omer Sajid Jeremy Zheng

  • Connection between TA and Mission Planner (MavProxy MavLink Forwarding)

  • Setup OTA link on Houston with RFD

  • Fly in 180° arc around TA (only testing yaw tracking capability)

  • If needed, possible Procedure is listed below (same as a past flight test)

...

TRACKING ANTENNA TEST PROCEDURE (HOUSTON)

Procedure

Goals / Objectives

Knockoff criteria

  1. Setup

LINK TO DOCUMENTATION FOR SETUP

  1. Hover altitude 5m in loiter.

  • Verify antenna is able to point towards the drone

< list of reasons why you would want to stop the test >

  1. Translation motion:

Move in periodic squares moving away from the tracking antenna.

  • Ensure that the tracking antenna remains pointed at the drone

  • tracking antenna loses track on the drone

    • Hover in place and wait.

Timeline

...

  • Batteries charging 9:30

  • Leave bay at 10:30

    • Drive or walk to CIF tbd

  • 10:50 begin Flight Card 13

    • Uninstall Lidar at the end of flight

  • 11:15 begin Flight Card 1

    • Flight 1

    • Flight 2

    • Flight 3

    • Flight 4

  • 11:40 50 begin Flight Card 32

    • Possible delay for installing IR camera for CV camera

    12:10

    • Flight 1

    • Flight 2

    • Flight 3

    • Flight 4

  • 12:25 begin Flight Card 4

  • 12:50 head back to bay

  • 12:30 debrief

  • 12:45 1:10 charge batteries to storage

  • 1:15 debrief and unpack flight test box

Attendees

Evan Janakievski

Sam Zhang Yuchen Lin

Maxwell LouOmer Sajid

Tochi Okoro

Vyomm Khanna

Ashish Agrahari

Andrew Shum

Key People Not Attending

If you’re directly involved in this stuff and cannot make this test please note it here. All onsite FTCs and subteam leads should fill this out please.

Debrief

Flight Card 1

Recap

Photos below are also in onedrive

Takeoff location

Landing pad to detectlocation Flights 1-3

Landing pad to detect location Flight 4

TakeoffLocation_pic1.pngImage Added

LandingPadToDetectLocation_Flight4_Pic1.pngImage Added

LandingPadToDetectLocation_Flight1_to_3_Pic1.pngImage Added

TakeoffLocation_pic2.pngImage Added

LandingPadToDetectLocation_Flight4_Pic2.pngImage Added

LandingPadToDetectLocation_Flight1_to_3_Pic2.pngImage Added

Flight 1 - 5 m altitude

West

Flight 2 - 10 m altitude

West

Flight 3 - 20 m altitude

West

Flight 4 - 20 m altitude

315 deg (i.e. in NW direction)

Good

  • Geolocation logs worked, will update with photos after checking them

  • Auto starts with Houston when powered up

Bad

  • Camera direction needs to be labeled for installing

    • Distance/offset for camera location needs to be known at some point

Flight Card 2

Recap

Cloudy weather

Grass in the background

IR emitter set on top of controller box

Flight 1 - 5 m

Flight 2 - 10 m

Flight 3 - 15-20 m

Good

  • Lots of images

  • Different altitudes

Bad

  • Not great images

    • Need a better camera with global shutter due to vibe and drifting of drone

  • IR camera broke before flight test, glued back on (Nathan glued it)

  • Took a while prior to leaving for flight to get set up on Pi

Flight Card 3

Recap

Flight 1 - completed script

Flight 2 - tested canceling script switching to loiter

Good

  • Script completed well, armed and took off drone in auto and successfully switched from Auto to Guided at flags

Bad

  • Could not switch from Auto/Guided to Loiter in the middle of the script

    • Was able to switch to RTL though

  • RTL switch to Loiter cut thrust and drone fell, caught with RTL but switch to loiter caused same issue

    • Hard landing, nothing was damaged

    • Thrust on controller was at 50% during this too

    • Test script on drone with lower altitude to problem solve switching to loiter issue

      • Simulation successfully switched from auto/guided to loiter, but did not translate to real flight

Action Items

  • Check photos for IR and Geolocation

  • Check path from guided mode and compare to simulation

  • Check RTL to loiter transition issues