Analyzing the architectural limits of centralized cloud schedulers versus distributed edge mesh computing in giant warehouses.
As fleet sizes cross 500 agents, pure centralized planners experience exponential solve-time scaling. MassSwarm solves this by partitioning the warehouse floor into autonomous spatial zones connected by boundary reservation ports.
Switching from full trajectory streaming to delta-compressed waypoint vectors reduces Wi-Fi bandwidth utilization from 120 Mbps to less than 4 Mbps across 1,000 robots.
In a 100,000 m² distribution center, the partitioned CBS-MAPF framework sustained 99.99% on-time order fulfillment during peak holiday volumes of 1.2M parcels/day.