---

ABOUT

The Git repository for the UMUAS (2023/2024) | Currently run by: Daniel Markuza

---

Design Plan:

Object Detection:

• Detect landing pad -> orange hexagon
• OpenCV
• Objectives:
  • Order the landing pad
  • Work session building model
  • Producing datasets to train the model
  • Configure Jetson Nano
    • Override Network Files (to be able to connect to the local host)
  • Set up training model on Jetson Nano
    • Connect FPV cameras
• Once done, create proper path to actually land at a landing pad (4 levels of difficulty)

Flight Termination:

• Detect loss of connection with GCS
• Land "safely"
  • Least damage possible
  • Not hurting anyone
• Integrate Python into C
• Set up switch between autopilot and normal pilot
  • Pixhawk 4
  • QGroundControl

MISC:

• Complete lap algorithm for task 1
• PID Tuning:
  • Adjust 3 variables for drone to fly and work properly
  • Variables in auto pilot code

---

Goals (per iteration):

Iteration 1 (INCOMPLETE):

• Object Detection:
  • Build the initial model
  • Produce datasets
  • Train model
• Flight Termination
  • Figure out how to detect loss of connection
  • Set up switch between the pilot and the autopilot
    • Talk with GCS about the button to switch between modes
  • Get started on figuring out how to get to drone to land safely

Iteration 2 (Deadline: March 27th):

• Iteration 1 revisions:
  • Communicate what we want changed
  • Revise deadlines
• Object Detection:
  • Negative images (ranging from super off to simialr to pad) - Kfir (Wednesday 13th)
  • Breaking videos into frames - Toni, Daniel (Saturday 9)
  • Generate dataset with area for the model focus on (manually or OpenCV) - Workshop (Saturday 9)
  • Build the model - Workshop (Wednesday, March 13th)
• Flight Termination:
  • Establish and allow connections on the drone - Workshop
  • Detect loss of connection with GCS - Workshop
  • Detect Critical Failure on drone - Workshop

Iteration 3 (Deadline: April 18th):

• Iteration 2 revisions:
  • Communicate what we want changed
  • Revise deadlines
• Object Detection:
  • Path algorithm for task 2
  • set up hardware to work with software
    • FPV cameras from drone working with software
    • Set it up on CPU on board the drone
• Flight Termination:
  • Run simulation
  • Land safely
    • find a safe location to "crash"
    • Just land (Systems intact just lost connection)
      • Fail safe
• MISC:
  • Algorithm for task 1:
    • Determine how many laps are possible (battery power, and etc.)

Iteration 3 (Deadline: N/A):

• PID Tunning
• Final testing
  • Make sure object detection model works with drone
  • Make sure algorithms work with drones
  • Make flight termination system