Datasets:
metro large_string | arm large_string | arm_name large_string | heights large_string | seed int64 | measured_hours float64 | conflicts_10m int64 | conflicts_15m int64 | conflicts_20m int64 | tight_15m_5m int64 | conflicts_10m_unbounded int64 | amazon_wing int64 | amazon_zipline int64 | wing_zipline int64 | flights int64 | sites_wing int64 | sites_zipline int64 | sites_amazon int64 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
dfw | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 1 | 2 | 41 | 85 | 136 | 18 | 43 | 5 | 35 | 1 | 13,103 | 75 | 75 | 6 |
dfw | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 2 | 2 | 39 | 73 | 101 | 20 | 40 | 6 | 31 | 2 | 12,808 | 75 | 75 | 6 |
dfw | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 3 | 2 | 50 | 84 | 130 | 20 | 50 | 13 | 37 | 0 | 13,025 | 75 | 75 | 6 |
dfw | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 4 | 2 | 39 | 75 | 115 | 14 | 41 | 7 | 32 | 0 | 13,084 | 75 | 75 | 6 |
dfw | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 5 | 2 | 60 | 99 | 147 | 25 | 61 | 7 | 50 | 3 | 13,175 | 75 | 75 | 6 |
dfw | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 6 | 2 | 37 | 78 | 120 | 16 | 37 | 12 | 22 | 3 | 13,017 | 75 | 75 | 6 |
dfw | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 7 | 2 | 31 | 78 | 123 | 10 | 32 | 11 | 18 | 2 | 13,136 | 75 | 75 | 6 |
dfw | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 8 | 2 | 53 | 102 | 145 | 25 | 55 | 9 | 41 | 3 | 13,014 | 75 | 75 | 6 |
dfw | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 9 | 2 | 60 | 111 | 161 | 31 | 63 | 14 | 44 | 2 | 13,068 | 75 | 75 | 6 |
dfw | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 10 | 2 | 34 | 77 | 129 | 16 | 37 | 4 | 29 | 1 | 12,998 | 75 | 75 | 6 |
dfw | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 1 | 2 | 13 | 49 | 50 | 2 | 14 | 0 | 4 | 9 | 13,103 | 75 | 75 | 6 |
dfw | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 2 | 2 | 6 | 40 | 44 | 1 | 6 | 1 | 3 | 2 | 12,808 | 75 | 75 | 6 |
dfw | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 3 | 2 | 11 | 48 | 52 | 0 | 12 | 5 | 6 | 0 | 13,025 | 75 | 75 | 6 |
dfw | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 4 | 2 | 5 | 41 | 49 | 1 | 6 | 2 | 2 | 1 | 13,084 | 75 | 75 | 6 |
dfw | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 5 | 2 | 6 | 61 | 68 | 3 | 6 | 1 | 3 | 2 | 13,175 | 75 | 75 | 6 |
dfw | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 6 | 2 | 10 | 43 | 51 | 4 | 10 | 2 | 3 | 5 | 13,017 | 75 | 75 | 6 |
dfw | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 7 | 2 | 11 | 30 | 38 | 0 | 11 | 2 | 6 | 3 | 13,136 | 75 | 75 | 6 |
dfw | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 8 | 2 | 7 | 53 | 58 | 2 | 7 | 2 | 3 | 2 | 13,014 | 75 | 75 | 6 |
dfw | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 9 | 2 | 15 | 61 | 69 | 2 | 15 | 6 | 5 | 4 | 13,068 | 75 | 75 | 6 |
dfw | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 10 | 2 | 7 | 43 | 52 | 1 | 7 | 1 | 3 | 3 | 12,998 | 75 | 75 | 6 |
dfw | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 1 | 2 | 8 | 52 | 100 | 4 | 8 | 4 | 3 | 1 | 13,103 | 75 | 75 | 6 |
dfw | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 2 | 2 | 9 | 44 | 71 | 4 | 9 | 3 | 4 | 2 | 12,808 | 75 | 75 | 6 |
dfw | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 3 | 2 | 15 | 54 | 99 | 0 | 16 | 12 | 3 | 0 | 13,025 | 75 | 75 | 6 |
dfw | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 4 | 2 | 7 | 46 | 88 | 4 | 8 | 5 | 2 | 0 | 13,084 | 75 | 75 | 6 |
dfw | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 5 | 2 | 15 | 50 | 99 | 6 | 15 | 9 | 3 | 3 | 13,175 | 75 | 75 | 6 |
dfw | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 6 | 2 | 16 | 54 | 93 | 4 | 16 | 10 | 3 | 3 | 13,017 | 75 | 75 | 6 |
dfw | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 7 | 2 | 16 | 67 | 112 | 4 | 16 | 8 | 6 | 2 | 13,136 | 75 | 75 | 6 |
dfw | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 8 | 2 | 21 | 61 | 108 | 7 | 21 | 15 | 3 | 3 | 13,014 | 75 | 75 | 6 |
dfw | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 9 | 2 | 18 | 71 | 120 | 9 | 18 | 11 | 5 | 2 | 13,068 | 75 | 75 | 6 |
dfw | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 10 | 2 | 14 | 55 | 101 | 3 | 14 | 6 | 7 | 1 | 12,998 | 75 | 75 | 6 |
dfw | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 1 | 2 | 33 | 82 | 136 | 11 | 35 | 6 | 26 | 1 | 13,103 | 75 | 75 | 6 |
dfw | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 2 | 2 | 23 | 71 | 98 | 8 | 24 | 3 | 18 | 2 | 12,808 | 75 | 75 | 6 |
dfw | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 3 | 2 | 35 | 86 | 133 | 1 | 35 | 16 | 19 | 0 | 13,025 | 75 | 75 | 6 |
dfw | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 4 | 2 | 30 | 78 | 119 | 8 | 31 | 6 | 24 | 0 | 13,084 | 75 | 75 | 6 |
dfw | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 5 | 2 | 46 | 97 | 148 | 15 | 47 | 10 | 33 | 3 | 13,175 | 75 | 75 | 6 |
dfw | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 6 | 2 | 33 | 78 | 119 | 6 | 33 | 13 | 17 | 3 | 13,017 | 75 | 75 | 6 |
dfw | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 7 | 2 | 17 | 76 | 127 | 6 | 17 | 9 | 6 | 2 | 13,136 | 75 | 75 | 6 |
dfw | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 8 | 2 | 49 | 98 | 147 | 12 | 51 | 13 | 33 | 3 | 13,014 | 75 | 75 | 6 |
dfw | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 9 | 2 | 46 | 109 | 162 | 14 | 47 | 13 | 31 | 2 | 13,068 | 75 | 75 | 6 |
dfw | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 10 | 2 | 31 | 81 | 125 | 4 | 32 | 9 | 21 | 1 | 12,998 | 75 | 75 | 6 |
houston | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 1 | 2 | 88 | 215 | 331 | 30 | 89 | 19 | 65 | 4 | 19,426 | 75 | 65 | 5 |
houston | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 2 | 2 | 123 | 263 | 355 | 43 | 126 | 16 | 101 | 6 | 19,091 | 75 | 65 | 5 |
houston | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 3 | 2 | 94 | 205 | 310 | 42 | 100 | 23 | 66 | 5 | 19,055 | 75 | 65 | 5 |
houston | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 4 | 2 | 83 | 171 | 260 | 38 | 88 | 15 | 60 | 8 | 19,288 | 75 | 65 | 5 |
houston | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 5 | 2 | 69 | 166 | 256 | 20 | 70 | 11 | 51 | 7 | 19,197 | 75 | 65 | 5 |
houston | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 6 | 2 | 114 | 242 | 353 | 51 | 121 | 15 | 91 | 8 | 19,167 | 75 | 65 | 5 |
houston | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 7 | 2 | 85 | 198 | 315 | 39 | 89 | 7 | 70 | 8 | 18,983 | 75 | 65 | 5 |
houston | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 8 | 2 | 88 | 201 | 305 | 30 | 93 | 5 | 79 | 4 | 19,234 | 75 | 65 | 5 |
houston | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 9 | 2 | 85 | 204 | 332 | 34 | 88 | 22 | 47 | 16 | 19,350 | 75 | 65 | 5 |
houston | A0 | omaha_profile | Amazon 61 m out / 105 m back (FAA Omaha noise analysis); Wing 46-91 m; Zipline 101 m | 10 | 2 | 86 | 191 | 289 | 34 | 86 | 18 | 59 | 9 | 19,179 | 75 | 65 | 5 |
houston | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 1 | 2 | 17 | 95 | 104 | 6 | 17 | 3 | 3 | 11 | 19,426 | 75 | 65 | 5 |
houston | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 2 | 2 | 36 | 131 | 140 | 10 | 36 | 4 | 22 | 10 | 19,091 | 75 | 65 | 5 |
houston | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 3 | 2 | 12 | 91 | 104 | 2 | 12 | 3 | 4 | 5 | 19,055 | 75 | 65 | 5 |
houston | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 4 | 2 | 17 | 82 | 98 | 6 | 21 | 4 | 8 | 5 | 19,288 | 75 | 65 | 5 |
houston | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 5 | 2 | 10 | 65 | 71 | 1 | 10 | 0 | 3 | 7 | 19,197 | 75 | 65 | 5 |
houston | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 6 | 2 | 24 | 118 | 128 | 5 | 27 | 6 | 12 | 6 | 19,167 | 75 | 65 | 5 |
houston | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 7 | 2 | 14 | 90 | 98 | 4 | 15 | 3 | 4 | 7 | 18,983 | 75 | 65 | 5 |
houston | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 8 | 2 | 18 | 94 | 103 | 7 | 19 | 4 | 7 | 7 | 19,234 | 75 | 65 | 5 |
houston | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 9 | 2 | 22 | 76 | 86 | 6 | 23 | 8 | 3 | 11 | 19,350 | 75 | 65 | 5 |
houston | A2 | shared_out | Each company its own height inside its filed band: Wing 46-71 m, Amazon 81 m out / 115 m back, Zipline 101 m | 10 | 2 | 22 | 82 | 93 | 8 | 22 | 3 | 8 | 11 | 19,179 | 75 | 65 | 5 |
houston | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 1 | 2 | 26 | 151 | 265 | 9 | 26 | 20 | 2 | 4 | 19,426 | 75 | 65 | 5 |
houston | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 2 | 2 | 36 | 189 | 284 | 12 | 38 | 14 | 16 | 6 | 19,091 | 75 | 65 | 5 |
houston | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 3 | 2 | 25 | 143 | 249 | 4 | 25 | 18 | 2 | 5 | 19,055 | 75 | 65 | 5 |
houston | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 4 | 2 | 29 | 122 | 207 | 3 | 33 | 16 | 5 | 8 | 19,288 | 75 | 65 | 5 |
houston | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 5 | 2 | 17 | 120 | 209 | 2 | 17 | 8 | 2 | 7 | 19,197 | 75 | 65 | 5 |
houston | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 6 | 2 | 32 | 155 | 272 | 7 | 34 | 14 | 10 | 8 | 19,167 | 75 | 65 | 5 |
houston | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 7 | 2 | 19 | 136 | 250 | 7 | 20 | 7 | 4 | 8 | 18,983 | 75 | 65 | 5 |
houston | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 8 | 2 | 14 | 129 | 234 | 2 | 15 | 2 | 8 | 4 | 19,234 | 75 | 65 | 5 |
houston | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 9 | 2 | 35 | 152 | 281 | 14 | 36 | 17 | 2 | 16 | 19,350 | 75 | 65 | 5 |
houston | R | high_return | Amazon 69 m out / 122 m back (secondhand staff remark only); Wing 46-91 m; Zipline 101 m | 10 | 2 | 36 | 141 | 237 | 9 | 36 | 20 | 7 | 9 | 19,179 | 75 | 65 | 5 |
houston | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 1 | 2 | 70 | 213 | 332 | 17 | 71 | 22 | 44 | 4 | 19,426 | 75 | 65 | 5 |
houston | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 2 | 2 | 81 | 263 | 366 | 18 | 85 | 14 | 61 | 6 | 19,091 | 75 | 65 | 5 |
houston | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 3 | 2 | 68 | 210 | 315 | 16 | 72 | 24 | 39 | 5 | 19,055 | 75 | 65 | 5 |
houston | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 4 | 2 | 64 | 175 | 263 | 16 | 69 | 19 | 37 | 8 | 19,288 | 75 | 65 | 5 |
houston | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 5 | 2 | 47 | 173 | 263 | 10 | 48 | 9 | 31 | 7 | 19,197 | 75 | 65 | 5 |
houston | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 6 | 2 | 89 | 237 | 359 | 19 | 94 | 19 | 62 | 8 | 19,167 | 75 | 65 | 5 |
houston | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 7 | 2 | 64 | 202 | 318 | 18 | 67 | 9 | 47 | 8 | 18,983 | 75 | 65 | 5 |
houston | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 8 | 2 | 65 | 199 | 310 | 12 | 68 | 3 | 58 | 4 | 19,234 | 75 | 65 | 5 |
houston | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 9 | 2 | 70 | 204 | 335 | 19 | 72 | 22 | 32 | 16 | 19,350 | 75 | 65 | 5 |
houston | S1 | best_sourced | Amazon 69 m out / back anywhere in 85-115 m (best-sourced); Wing 46-91 m; Zipline 101 m | 10 | 2 | 74 | 193 | 289 | 14 | 74 | 20 | 45 | 9 | 19,179 | 75 | 65 | 5 |
Shared Sky: delivery drones sharing low-altitude airspace
Dataset version 1.0, 21 November 2026.
Wing, Zipline and Amazon have filed with the FAA to fly delivery drones over the same parts of Houston and Dallas–Fort Worth. This dataset comes from a pre-registered simulation that flies their planned operations together, with no coordination between the companies, and counts planned conflicts: moments when two different companies' planned paths come within 30 m sideways and less than 10 m in height. Meetings between drones of the same company are not counted: each company is assumed to keep its own flights apart.
Headline result: giving each company its own height inside the band it already filed cuts planned conflicts by 72 % in Houston (95 % interval 67–76 %) and 73 % in Dallas–Fort Worth (66–80 %), against Amazon at its best-sourced heights. You can recompute both numbers from headline_runs in two lines (below).
- 3D site: https://forgottenl5.github.io/shared-sky
- Code, pre-registrations and every result (77 registered studies; 316 predictions passed, 97 failed, 4 not scored): https://github.com/forgottenl5/shared-sky
- Pre-registration for the headline study:
planned/HEADLINE2_PREREG.md, committed with its runner and analysis before any run. Results:planned/HEADLINE2_RESULTS_2026-10-25.md
Quick start
from datasets import load_dataset
runs = load_dataset("forgottenl5/shared-sky", "headline_runs", split="train").to_pandas()
# Recompute the headline: shared-out heights (A2) vs best-sourced Amazon (S1), 10 m vertical rule
for metro, d in runs.groupby("metro"):
cut = 1 - d[d.arm == "A2"].conflicts_10m.sum() / d[d.arm == "S1"].conflicts_10m.sum()
print(metro, f"{cut:.0%} fewer planned conflicts")
# dfw 73% fewer planned conflicts
# houston 72% fewer planned conflicts
Without the datasets library:
import pandas as pd
runs = pd.read_parquet("hf://datasets/forgottenl5/shared-sky/headline_runs.parquet")
Plot where planned conflicts happen in the site's demo patch:
import matplotlib.pyplot as plt
c = load_dataset("forgottenl5/shared-sky", "demo_conflicts", split="train").to_pandas()
c = c[(c.metro == "houston") & (c.kind == "conflict")]
fig, axes = plt.subplots(1, 2, figsize=(10, 5), sharex=True, sharey=True)
for ax, sc in zip(axes, ["sourced", "shared"]):
d = c[c.scenario == sc]
ax.scatter(d.x_m, d.y_m, c=d.z_m, s=12, cmap="viridis")
ax.set_title(f"{sc}: {len(d)} planned conflicts in 2 h")
plt.show()
What's in it
| Config | Rows | One row is |
|---|---|---|
headline_runs (default) |
80 | one run of the headline study: 2 metros × 4 height rules × 10 seeds, 2 measured hours each, on the real filed layouts |
demo_scenarios |
12 | one scenario of the website's demo patch (6 scenarios × 2 metros), with its counts and a plain-language description |
demo_conflicts |
563 | one planned conflict (or tight meeting) in the demo patch: time, place, height, company pair |
demo_flights |
67,221 | one planned delivery flight in the demo patch: company and its 3D path as keyframes |
headline_runs
The study behind the headline number. Amazon sites are geocoded from its filings; Wing and Zipline sites are placed at random per seed in the filed counts (Houston: 75 Wing, 65 Zipline, 5 Amazon; DFW: 75 / 75 / 6).
| Column | Meaning |
|---|---|
metro |
houston or dfw |
arm, arm_name, heights |
height rule: A0 Omaha profile, S1 best-sourced Amazon, R high-return case (secondhand staff remark only), A2 shared out |
seed |
1–10 (Wing/Zipline layout and traffic) |
measured_hours |
2 (after a 1 h warm-up) |
conflicts_10m |
planned conflicts between companies: within 30 m sideways and < 10 m vertically (the headline rule) |
conflicts_15m, conflicts_20m |
the same with 15 m or 20 m vertical |
tight_15m_5m |
within 15 m sideways and < 5 m vertically |
conflicts_10m_unbounded |
10 m count without bounding to the measured window (a fix checked in the study: 3–4 % higher in the A0 arm, 2.9–4.1 % across all arms) |
amazon_wing, amazon_zipline, wing_zipline |
conflicts_10m split by company pair |
flights, sites_* |
flights flown and sites per company |
demo_scenarios, demo_conflicts, demo_flights
What the 3D website plays: one seed on a 12 km × 12 km patch with sites spaced every 3 km (a synthetic layout, not real store locations), with each metro's company mix. The patch is denser than the real layouts, so its counts differ from the headline (Houston best-sourced → shared out: 92 → 33, a 64 % cut).
Scenarios: filed, sourced, reported, shared (height rules as above); checked and layered (shared out plus pre-flight plan sharing). Each row of demo_scenarios has the full description. The website's three "bad day" scenarios (1 in 5 drones hovers mid-route) are left out of version 1.0: a review found that the engine undercounts conflicts for flights that change their plan in flight, and they return once that is fixed and re-checked.
- Coordinates
x_m(east) andy_m(north) are metres from the centre of the patch;z_mis height above flat ground in metres (no terrain or buildings). Flights start at sites in and around the patch and can fly well beyond it; only meetings with at least one drone inside the 12 km zone are counted. - Times
t_sare seconds from the start of the 2-hour measured window (negative = before it). demo_flights.t_s/x_m/y_m/z_mare keyframes: the drone flies in straight lines between them.flightnumbers are per scenario and do not match across scenarios.demo_conflicts.kind:conflict(30 m / 10 m rule) ortight(15 m / 5 m, recorded as its own row). Each row is one meeting of two flights; rows can share a time and place.x_m,y_mare the position of the first drone of the pair (within about 30 m of the zone when that drone is just outside it);z_mis the mean height of the two drones.demo_scenarios.conflicts_10mcounts the 2-hour measured window (what the site shows);conflicts_10m_unboundedalso counts meetings after the window ends, as an older counting method did.
How it was made
- Inputs: the companies' FAA environmental filings (site counts, flight caps, heights and noise profiles), Amazon's stated Richardson average height, a secondhand homeowners' association note for the high-return case, 2025 Open-Meteo weather for other studies. Every source is listed in the repository.
- Engine: planned flight paths only; no avoidance and no coordination (except the
checked/layeredscenarios). Conflicts are checked 4 times a second. - Statistics (headline): 95 % percentile bootstrap over the 10 seeds, paired by seed (10,000 resamples).
- Export:
tools/export_dataset.pyin the repository rebuilds every file here from the public release and checks the 72 % / 73 % cuts.
Limitations (read before using)
- This is simulated traffic, not real flights. A planned conflict is not a collision or a near-miss: real operators already share flight plans before take-off (ASTM F3548), so the no-coordination baseline is a reference point for how much each fix removes, not a picture of today's risk.
- Amazon's return height is not public. If it is 400 ft (the high-return case), the cut is 29 % (Houston) and 35 % (DFW), not 72 % / 73 %.
- How much vertical room companies need decides the number: if 15 m is required, the cut is 55 % / 45 %.
- Wing and Zipline sites are random in the headline study; filed caps on overlapping service areas are not applied.
- The demo tables are one seed on a dense synthetic patch, made for the website. Use
headline_runsfor any claim about the study. - Failed predictions: 97 of the project's pre-registered predictions failed, including one in this study (H3: the cut at 15 m stayed above 50 % in Houston). They are reported as failures in the repository.
- Not affiliated with Wing, Zipline, Amazon or the FAA. Not a safety assessment.
License
CC BY 4.0. Credit: "Shared Sky (github.com/forgottenl5/shared-sky)". The tables contain no OpenStreetMap, ADS-B or personal data. The simulation code is AGPL-3.0-or-later (in the repository).
Citation
@misc{forgottenl5_2026_sharedsky,
author = {forgottenl5},
title = {Shared Sky: a pre-registered simulation of delivery drones and air taxis over Houston and Dallas--Fort Worth},
year = {2026},
howpublished = {\url{https://github.com/forgottenl5/shared-sky}},
note = {Dataset version 1.0, https://huggingface.co/datasets/forgottenl5/shared-sky}
}
- Downloads last month
- 2