ChromaVertex Project Journal

Project Outline

Hardware Stream

• VertexBot PCB design

Software Stream

1. AprilTag processing with ESP32-S3-CAM (AprilTags are in fixed, known positions) –> learn onboard vision and localization
   • Test detection range, determine optimal AprilTag dimensions
   • Test detection robustness (e.g. occlusion)
   • Test simultaneous detection of multiple tags
   • Test detection while rotating
2. VertexBots + AprilTag beacons at fixed positions + wheel odometry –> learn dead reckoning
3. VertexBots + AprilTag beacons at fixed positions + wheel odometry + IMU –> learn sensor fusion
   • Test how long pose estimation stays “sufficiently” accurate before re-calibration with AprilTag beacons
4. VertexBots + onboard RGB LED identification + wheel odometry + IMU
   • Design decisions need to made on the geometry of the RGB LED setup (ring? pillar?)
   • Validate robot-to-robot perception against ground truth (i.e. AprilTag beacons at fixed locations)
5. Cooperative localization –> learn distributed perception and pose estimation
   • Robots detect one another
   • Exchange relative pose observations
   • Use triangulation/trilateration and distributed algorithms to estimate poses
6. Advanced behaviours: synchronization, formation control, collective decision making and swarm intelligence
   • Boids algorithm is a good starting point
   • Simultaneous Localization and Mapping (SLAM)

The Journal

1-June-2026

Hardware design discussions begin:

• A rough idea about the VertexBot structure emerged: We will use two back-to-back ESP32-S3-CAM boards and wide-angle lenses to cover close to 360 degrees.
• Patrick started sketching the main PCB. We are waiting for the arrival of the first ESP32-S3-CAM modules to test.
• In the July-2026 monthly meeting we will experiment with AprilTag identification algorithms.

7-July-2026

We are thinking about how to localize the robots with minimal effort. I am experimenting with the ESP AprilTag library. So far my program is not successful detecting any AprilTags :-(

Meanwhile, Mick has created a light pillar that we could use for localizing VertexBots in two ways:

• We can have stationary “beacons” at the corners of an arena that vertexbots detect them to run triangulation to estimate their poses, then
• We can have “light towers” on each VertexBot for driving relative pose estimation algorithms.

Patrick and I have almost finalized the pin assignment of the ESP32-S3-CAMs.