<!-- Generated by scripts/sync-agent-content.mjs. Edit the shared public sources, then regenerate. -->
# MOSAIC

Robot fleet coordination and spatial debugging through shared telemetry.

- Dates: 2025–2026
- Status: Lab platform prototype
- Areas: Software engineering; Robotics
- Stack: Python · ROS 2 · OptiTrack · NDI
- Topics: Python; ROS; Robotics; Distributed Systems; UDP; Autonomy; Visualization; Physical Computing; Research
- Public case study: https://mgiorgetti.com/project_details/project-mosaic.html

## Problem

A robot fleet needs coordinated operation and a shared view of telemetry in physical lab space, with movement authority kept separate from visualization.

## Contribution

Robotics systems engineering across fleet coordination, ROS 2 robot agents, operator tools, telemetry contracts, calibrated floor visualization, recording and observation-only replay.

## Technical approach and public case study

Coordinate a robot fleet and make its behavior visible in physical space, with robot-local movement control and calibrated floor visualization.

- AI & Robotics Lab project

- Python · ROS 2 · OptiTrack

- UDP · NDI · TouchDesigner

Paths and manual commands

Commands in · fleet state out

Sensors & pose → path following → motor commands

High-level commands travel through the supervisor; motor commands are published by each robot agent.

### Control responsibilities

The supervisor handles coordination. Each robot reads its own sensors and OptiTrack pose, executes movement, and checks whether that pose is still current.

**Robot agent**

Owns path following, manual drive, stopping decisions, and the only local cmd_vel publisher.

**Supervisor bridge**

Translates UDP commands into ROS 2 actions and aggregates normalized robot state.

**Operator tools**

The Tk interface issues commands. Gama receives fleet state through a separate UDP export.

### Shared topics and fleet configuration

#### Normalized ROS topics

Agents publish MOSAIC status, pose, diagnostics, and path progress. The supervisor consumes these interfaces instead of depending on vendor-specific robot topics.

#### Shared fleet configuration

A common configuration defines robot namespaces, endpoints, and device roles. Launch packages and deployment scripts keep startup consistent across machines.

### Put observations back into physical space

MOSAIC’s visualization component maps selected robot telemetry and independent tracking data onto a calibrated floor display. Source timestamps, units, coordinate frames, and provenance stay attached to observations; missing or stale evidence is shown as unavailable.

- Observe Read selected robot telemetry and independent tracking data.

- Check Retain the timestamp, coordinate frame, and validity of each observation.

- Project Map observations through shared calibration into the floor display.

### Inspect a run without commanding robots

Fleet control and visualization are components of the same project with separate responsibilities. The visualization hub owns calibration between tracking coordinates and the projection surface. Movement and stopping authority remain with each robot agent.

**Independent evidence**

External tracking and robot-reported reasoning remain distinguishable in the display.

**Versioned recordings**

Typed observations and metadata preserve the timing context of a debugging session.

**Observation-only replay**

Recorded evidence feeds the display and checks without sending movement commands to live robots.

### From Hydra to robot-local control

The linked public repository documents the earlier Hydra controller: a centralized ROS 1 backend and ROS 2 robot client. The current architecture described here moves path execution and stopping decisions into ROS 2 agents on each robot.

The public repository is historical source for the fleet controller’s evolution. The visualization and replay component is described separately in the visualization & debugging detail.

### Physical validation requirements

Fleet membership is configured in advance. Validating movement depends on the robots, their ROS interfaces, OptiTrack, and the lab network. The visualization component also requires projector alignment, tracking calibration, and checks on the lab hardware.

## Results and limitations

MOSAIC is one project with fleet-control and visualization/debugging components. Robot agents own movement execution and stopping; the visualization observes and projects telemetry. The linked public repository documents the earlier Hydra architecture. Current physical validation depends on configured robots, ROS interfaces, OptiTrack, projector alignment and the lab network.

## Public supporting links

- [Portfolio case study](https://mgiorgetti.com/project_details/project-mosaic.html)
- [Historical Hydra source](https://github.com/Inventure71/SwarmProjectV1)
- [Visualization & debugging](https://mgiorgetti.com/project_details/project-mosaic-floor.html)
- [Matteo Giorgetti profile](https://mgiorgetti.com/profile.md)
