A sharp capacity limit for low-altitude urban airspace and its management implications

Low-altitude urban airspace is rapidly emerging as a metropolitan transport layer, yet the traffic density that shared airspace can safely sustain under decentralised local avoidance remains unknown. We address this question through agent-based simulations of three representative local-avoidance implementations: reactive velocity selection, force-based summation and constrained predictive search. All three deteriorate sharply within a common tested density window of φ ≈ 0.25–0.30. We define φ c r i t i s o ≈ 0.27 as the representative benchmark under the baseline isotropic protocol and safe completion rate (SafeCR) criterion. The results associate the transition with velocity-space crowding: increasing neighbour counts restrict feasible motion, with direct blocked-action diagnostics for the VO-filter and MPC-style implementations. An uncalibrated isotropic mean-field construction predicts a scale approximately 5% below the benchmark, and a simulation-calibrated empirical closure accounts for the residual. Directional-organisation tests with the VO-filter show that relative kinematics and blocked-direction union geometry shift outcomes within the transition window. We then evaluate two system-level responses. On the tested density grid, the bounded two-layer transition midpoint lies near 0.48, corresponding to an approximate 80% extension relative to the representative isotropic benchmark. Admission metering caps airborne occupancy by holding excess demand on the ground. These boundaries define an operating-regime map used to screen nine illustrative scenarios from Shenzhen and Paris under stated separation and extrapolation assumptions. The framework links local self-separation limits to system-level capacity management and provides a quantitative basis for evaluating structural and entry-control responses as demand grows.