Week 2

in Blog

GSOC Coding Week 2 Progress Report

Week 2: Writing and Testing a New Algorithm

After understanding the legacy structure last week, I focused on building and validating a new algorithm to test the current simulation setup. The original MyAlgorithm.py script was non-functional, so I had to write one from scratch to validate if the setup pipeline—including the perception and manipulation systems—was intact.

What I Did This Week

  • Wrote a new algorithm for the pick-and-place task using point cloud-based object detection
  • Tested the algorithm in the legacy Docker environment
  • Confirmed that the simulation and robot control flow still work under the new logic
  • Disabled constraints in the “place” function because the trajectory was constantly failing under the original settings
  • Recorded a working demo — note that the first pick-and-place action has a slightly weird motion due to disabled constraints

New test with working algorithm

Known Issues

Even though the algorithm now runs successfully, there are still a few remaining challenges:

  • The “place” trajectory fails when constraints are enabled — possibly due to unreachable poses or tight tolerances
  • The Gazebo world is still incomplete — some models or markers fail to spawn

Next Steps

Looking ahead, I plan to:

  • Fix the Gazebo environment so that all required models and objects load consistently
  • Start the ROS2 migration — set up the ROS2 workspace and begin porting the robot and camera pipeline
  • Evaluate better constraint strategies for “place” poses

📍 Posted from Barcelona, Spain
🧠 Project: Migration and Enhancement of Machine Vision Exercise to ROS2 + MoveIt2