- Shared map
- How many robots' maps are combined into one team map.
- Robots agree within
- Average distance between where different robots place the same object. Lower means the maps line up; optimizing the team map reduces it.
- Accuracy vs. truth
- How far the team map is from the true layout. Only knowable in a simulation.
- Data shared
- Bytes robots sent each other: object descriptors and optimization messages, versus megabytes for raw images.
- Agents
- The brain each robot uses: a built-in planner, or an open LLM running in your browser.