Multi-robot authoring—using several robots to complete a drawing—can cut completion time, but how much depends on how the drawing is decomposed and assigned. In a new arXiv preprint, Marowski and Min represent drawings as geometric stroke graphs and compare two allocation strategies: giving each connected component to a single robot, or decomposing components so multiple robots share the workload.
Simulating 385 drawings with teams of one to eight robots, the authors found that the shared-token pipeline achieved larger mean speedups relative to its own single-robot baseline. It also showed lower idle-time shares at larger team sizes. However, shared-token execution did not reduce completion time for every drawing, indicating that the benefit is not universal.
The results also point to task structure as a key factor. Differences in scaling behavior were associated with structural attribute groups, and those groups separated more clearly as team size grew. The authors argue that within-component workload and graph structure deserve attention alongside simple component counts when designing multi-robot authoring systems.