2024-11-01 Autonomy Meeting

Pathing

Jane Zeng

• Circle mode done yet?

• Waypoint circles?

  • PR is up, waiting for local_global_conversion to move to common

Airside

Ashish Agrahari

• Cluster estimation progress?

• Geolocation/cluster estimation output to KML module and worker: call it KML processing worker?

  • Also needs to be able to send data to ground through mavlink (can we send custom messages in mavlink?)

  • The ground will then run a script to parse the GPS coords and build KML file

Tasks:

• Achita Anantachina Classical CV method tests

  • R D to look into issue

• Shahvir Wahab Research ROS with ardupilot

  • Done

  • Testing - simulation of drone inputs (camera, GPS, etc.) VS SITL?

  • Tools? how they benefit CV/ML inferencing

    • either develop faster or runs faster

• Shahvir Wahab Testing ROS with ardupilot

  • Will start soon

• Victor Terme Auto landing module

  • Needs to talk to someone about the task

  • He will talk to Tong Zhang later

• Andrew Xu Logging for all workers

  • Working on logs

  • In progress

• Siddh Patel Create log merging script

  • Waiting for PR approval

• Rudimentary geolocation progress?

ML

Vibhinn Gautam

• No notes as of now.

IMACS

Balaji Leninrajan

• PRs for

  • Create widget to execute terminal scripts

  • Implement UI to change address and port

  • Create RTL (also reviewed)

• New tasks related to SITL setup soon

Obstacle Avoidance

Vyomm Khanna

• Implementation of VFH algorithm for obstacle avoidance coming along well.

• Updated algorithm to process LiDAR readings as a continuous stream instead of a file, display polar histogram after each oscillation (forward and backward)

• Updated algorithm to output path deflection angle (temporary waypoint direction) based on appropriate candidate valley identified.

• Wrapping up unit tests, and tweaking formula based on results.

• Scheduled to work on adding temp waypoints after this

IR Camera + detection

• IR camera has arrived and it works quite well.

  • The infrared emitter looks like a giant white/bright spot

  • Detection of bright/white spots is needed

  • We will be gathering a lot of data hopefully. Ideally we schedule a couple more flight tests for this purpose. # of images needed - target 1100 for now?

  • We will use classical CV for the model. Flight test on Saturday (and resulting images) will help determine which technique to implement.