Tower vs. Self-Organized Droneport ATC: What a 9-Cell Factorial Study Found

Every urban air mobility operator faces the same question before they build their first vertiport: do you put a centralized tower in charge, or let drones negotiate landing slots peer-to-peer? The answer determines infrastructure cost, safety margins, and how the system degrades when traffic spikes.

We ran a nine-cell factorial study to find out. Three factors, nine combinations, 30 seeds per cell across five traffic loads. The headline: self-organized coordination with ADS-B matches tower throughput below roughly 20–25 ops/hour — then falls behind sharply as load increases. And drones flying silent (no ADS-B, no tower) exceed safe separation thresholds at just 12 ops/hour.

The three factors

The study crosses coordination authority (centralized tower vs. fully decentralized self-organization), communications regime (continuous telemetry vs. terminal-only — silent during cruise), and observation modality (ADS-B cooperative broadcast, camera-only, sensor fusion, or nothing). Nine of the possible sixteen combinations are operationally meaningful; those are the cells we ran.

E2 — tower, continuous, ADS-B — is the control. E9 — self-organized, terminal-only, no broadcast — is the stress floor. Everything else lives between them on the throughput–safety Pareto frontier.

Reference cell: tower + ADS-B + continuous

The E2 reference cell at light load (5 ops/hour) achieves 140 ± 28 ops/hour completed throughput, zero loss-of-separation events, and 12.6 m minimum pairwise separation at 0.078 messages/second. The tower consistently clears arriving drones before any queuing builds. This is the ceiling.

Where self-organization works (and where it doesn't)

Self-organized coordination with ADS-B (E6) matches E2 at light load. Drones broadcast intent, negotiate pad claims by priority, and clear the holding circuit without central authority. The distributed protocol handles low-density traffic well.

Above the crossover load (~20–25 ops/hour for a six-pad configuration), the gap opens. The tower's global slot optimizer prevents approach conflicts that the distributed back-off protocol can only resolve by delaying one drone — compounding under sustained load. Self-org without broadcast (E7, E9) is worse: without visibility of neighbors mid-route, separation events rise steeply past 12 ops/hour.

The silent-cruise trap

Terminal-only communications (drones broadcast only in the approach and departure zones, silent during cruise) is attractive for bandwidth. At 5 ops/hour the cost is small. At 40 ops/hour, the absence of mid-air deconfliction triggers measurable LoS increases. Self-org + terminal + no broadcast (E9) — the configuration that costs the least to deploy — is also the most dangerous at commercial load targets.

What this means for vertiport design

The Pareto-efficient designs for high-density droneports all share two properties: centralized pad scheduling and cooperative surveillance (ADS-B or fusion). Self-organization is the right architecture for low-density corridors where infrastructure cost dominates. The crossover point — roughly 20 ops/hour for six pads — is the design threshold operators should use when choosing between the two architectures.

The broadcast necessity threshold (LoS exceeds 0.01 events/op around 12 ops/hour in no-broadcast cells) maps directly to a regulatory question: at what traffic density does Remote ID / ADS-B become mandatory? This study argues the answer is well below commercial delivery density, which is where most UAM operators intend to operate.

Full experimental design, metrics, and simulation methodology: research section. The kinematic harness and all nine cells are available at the Astral GitHub.