| 研究課題/領域番号 |
18H05925
|
|---|
| 配分区分 | 補助金 |
|---|
| 研究機関 | 国立研究開発法人宇宙航空研究開発機構 |
|---|
研究代表者 |
アンドレエバ森 アドリアナ 国立研究開発法人宇宙航空研究開発機構, 航空技術部門, 研究開発員 (30747499)
|
|---|
| 研究期間 (年度) |
2018-08-24 – 2020-03-31
|
| キーワード | ground holding / airborne delay / ground delay / machine learning / capacity loss |
|---|
| 研究実績の概要 |
A dual-component ground holding (GH) algorithm based on real-time air traffic classification and offline ground holding program parameter optimization is developed. The basic design principles governing the ground holding algorithm are developed. Numerical simulations are developed to quantitatively evaluate this new concept. GH program performance is evaluated based on airborne delay, ground delay, and lost throughput costs. Preliminary results show that the developed machine-learning-based traffic pattern classifier can propose ground holding control parameters which would result in 88% of the potential optimal savings.Besides, departure time delays are modeled and will be implemented to increase the fidelity of the developed algorithm.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
Numerical simulation to quantitatively evaluate each ground holding program parameters has been developed and used to create the database needed as input for the machine learning algorithm. Suppor vector machine has been applied as a preliminary machine learning algorithm candidate.Results has shown the potential of the proposed dual-component algorithm and also identified areas where work is needed, especially to increase fidelity. In particular, departure time errors have been modeled, and further improvements are going to be made in the second year.
|
| 今後の研究の推進方策 |
Research development will continue in two major directions 1) increase the fidelity by implementing more detailed and accurate flight time error models and departure time error models, and 2) investigate other machine larning techniques to increase the prediction accuracy, as well as air traffic controller's acceptance.
|
