Occupancy-Interval Grids for Multi-UAV Conflict Detection and Resolution in Urban Low-Altitude Airspace

Pre-departure UTM checks for urban logistics UAVs use a working UAV–UAV miss distance of 30 m in this study, together with façade clearance. Both occupants are stored on one occupancy-interval grid: DDA traversal writes [tin, tout) along each issued polyline, and static buildings occupy the same keys for the horizon. Interval overlap after an 8-neighbour halo, plus a cell-to-façade test, is only a candidate. A candidate is confirmed if continuous closest-point-of-approach (CPA) is below 30 m, or if the track violates a 5 m façade-column clearance on the 5 m height raster. OG-CDR* repairs that confirmed set—local SIPP while the load is moderate, otherwise full-fleet prioritised SIPP—and refuses leftover CPA hits. Tests use four public city tiles (The Hague, Rotterdam Delfshaven, Montreal VM05, and a 1.5 km New York crop). Halo-free DDA misses about 30% of 30 m CPA pairs. When Δx = dsep, the halo candidate set contains those positives; the same CPA and column tests then drop the false candidates. For Δx < dsep, the one-cell halo is empirical. Occupancy-only repair can empty the table and still leave CPA pairs; OG-CDR* clears the residual and accepts fewer flights. The resolver is not compared with complete MAPF search.