Hasnain Ali

Hasnain Ali

Research Fellow in the area of applied Machine Learning and Artificial Intelligence for Intelligent Transportation Systems.

Publications

A Deep Reinforcement Learning Approach for Airport Departure Metering under Spatial-Temporal Airside Interactions

Published in IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, 2022

We have developed novel machine learning algorithms to minimize airport airside congestion by intelligently controlling aircraft pushbacks. Empirical results have demonstrated that machine learning based control models can learn effective strategies to manage airside congestion by regulating the spatial-temporal evolution of airport traffic. Benefits of up to 44% reduction in taxi delays were achieved in Singapore Changi Airport that correspond to 2-minute savings in taxi-out time per aircraft. The proposed method has the potential to reduce airport emissions and carbon footprints alongside minimizing the airport congestion problem.

Recommended citation: Ali, H., Pham, D. T., Alam, S., & Schultz, M. (2022). A Deep Reinforcement Learning Approach for Airport Departure Metering Under Spatial–Temporal Airside Interactions. IEEE Transactions on Intelligent Transportation Systems. https://ieeexplore.ieee.org/document/9906869

Integrated Airside Landside Framework to Assess Passenger Missed Connections with Airport Departure Metering

Published in 10th International Conference on Air Transportation, 2022

Although DM implementation can contain airside congestion, it may adversely impact scheduled gate assignments. Using an integrated landside airside framework, this study aimed at evaluating the impact of DM on flight and passenger connections. Using Singapore Changi airport A-SMGCS data, as a case study, it is found that DM may transfers as much as 12 minutes of waiting time from taxiways to gates. This leads to increased gate conflicts which additionally delays the gate-in time of an arriving aircraft by 2-7 minutes on average. Gate reassignments were found to significantly reduce both the number of gate conflicts and the time an arrival must wait on the apron before parking at the gate. Through extensive simulations, given the existing uncertainties at Singapore Changi airport terminal 4, an MCT of 70 minutes is found sufficient to reduce the probability of missed connections. As a next step, connections aware departure metering policies will be developed for Singapore Changi airport with an aim to optimize the trade-off between taxi-delays (on airside) and missed connections (on landside). Moreover, impact of operational constraints like emergency landings, ground delay program etc., on passenger connectivity shall also be investigated.

Recommended citation: Ali, H., Pham, D. T., Alam, S., & Schultz, M. (2022). Integrated airside landside framework to assess passenger missed connections with airport departure metering. https://dr.ntu.edu.sg/handle/10356/160033

Deep Reinforcement Learning Based Airport Departure Metering

Published in 24th IEEE International Intelligent Transportation Systems Conference, 2021

In this work, we proposed a DRL approach for controlling pushback times (TSATs) of departures. We employed PPO algorithm to train DRL agent into finding an effective DM policy. Since multiple agents are designed to share a common, single-agent policy, the final learned policy can be copied onto any number of agents. Simulating airside surface traffic in a representative airport environment, it is shown that DRL approach is able to learn, on simulated scenarios, an effective DM policy to contain congestion on an airport airside. As a result, taxi delays and average taxi times are reduced significantly. Through extensive simula- tion experiments, it is demonstrated that the final trained policy naturally scales to various fleet sizes. With increase in traffic density, taxi time savings obtained by the DM policy improves without significant increase in average gate holding times.

Recommended citation: Ali, H., Thinh, P. D., & Alam, S. (2021, September). Deep Reinforcement Learning Based Airport Departure Metering. In 2021 IEEE International Intelligent Transportation Systems Conference (ITSC) (pp. 366-371). IEEE. https://ieeexplore.ieee.org/abstract/document/9565012

Dynamic Hot Spot Prediction by Learning Spatial-Temporal Utilization of Taxiway Intersections

Published in International Conference on Artificial Intelligence and Data Analytics for Air Transportation, 2020

In this work, we have predicted dynamic congestion hotspots on airside taxiway network to prevent safety incidents and delays. Hot spots- areas where multiple aircraft may come in close vicinity on taxiways, act as pre-cursors to airside conflicts. We used airside infrastructure and surveillance data of Changi airport to model aircraft arrival at different taxiway intersections both in temporal and spatial dimensions. The statistically learnt spatial-temporal model was then used to compute conflict probability at identified intersections. These hot spots were then visually displayed on the aerodrome diagram for heightened attention of air traffic controllers (ATCOs). In the opinion of Ground Movement ATCO, highlighted Hot Spots lead to better understanding of taxiway movements and increased situational awareness.

Recommended citation: Ali, H., Delair, R., Pham, D. T., Alam, S., & Schultz, M. (2020, February). Dynamic hot spot prediction by learning spatial-temporal utilization of taxiway intersections. In 2020 International Conference on Artificial Intelligence and Data Analytics for Air Transportation (AIDA-AT) (pp. 1-10). IEEE. https://ieeexplore.ieee.org/document/9049186

A Passenger-centric Model for Reducing Missed Connections at Low Cost Airports with Gates Reassignment

Published in IEEE Access, 2019

The goal of this work is to minimize missed connections due to arrival delays at airports. I quantified the impact of operational uncertainties on passenger connections using Singapore Changi International Airport Terminal 4 layout and six month ADS-B data. The proposed passenger centric model, in this study, also advises about reassignment of gates in a delay disrupted scenario. I found that missed connections can be significantly reduced by operationally maintaining higher turnaround time and minimum connection time at airports.

Recommended citation: Ali, H., Guleria, Y., Alam, S., & Schultz, M. (2019). A passenger-centric model for reducing missed connections at low cost airports with gates reassignment. IEEE Access, 7, 179429-179444. https://ieeexplore.ieee.org/document/8918374

Impact of Stochastic Delays, Turnaround Time and Connection Time on Missed Connections at Low Cost Airports

Published in Air Traffic Management R&D Seminar, 2019

The goal of this work is to minimize missed connections due to arrival delays at airports. I quantified the impact of operational uncertainties on passenger connections using Singapore Changi International Airport Terminal 4 layout and six month ADS-B data. ADS-B technology relies upon aircraft avionics, a constellation of GPS satellites, and a network of ground stations. In it, aircraft position is first determined by on-board satellite navigation systems. This information is then broadcasted by the ADS-B transponder (attached with aircraft) to be picked up by ADSB receivers on the ground. ADS-B data of aircraft movements (13,812 departures and 13,403 arrivals) to and from Changi airport was analysed in this study. The analysis helped to learn and thereafter simulate gate arrival and departure delays. By maintaining the flight turnaround time at 50 min, minimum connection time at 60 min and by containing arrival delays within 70% of the current delay spread at Terminal 4, transfer passenger missed connections can be prevented for almost all flights.

Recommended citation: Ali, H., Guleria, Y., Alam, S., Duong, V. N., & Schultz, M. (2019). Impact of stochastic delays, turnaround time and connection time on missed connections at low cost airports. In Proc. 13th USA/Eur. Air Traffic Management R&D Seminar. https://drive.google.com/file/d/1sdfxdCS-vtj_LaZINbVR03Lbi9xqrpb8/view

Discriminant Analysis using Ant Colony Optimization – An Intra-Algorithm Exploration

Published in International Conference on Computational Intelligence and Data Science, 2018

Classification is a data mining function which assigns input values to two or more designated classesof output, with the goal of accurately predicting target class for new input data. Discriminant analysis which falls under the broad gambit of classification data mining, is the statistical mapping of data values to one of the two predefined groups. In this study, the application of Ant Colony Optimizationto perform discriminant analysis has been explored. Predictive classification accuracy ofeighty-six percent has been achieved on an ordinal data set for the combination of one thousand ants and twenty-five instances (minimum) covered by one rule.

Recommended citation: Ali, H., & Kar, A. K. (2018). Discriminant analysis using ant colony optimization–an intra-algorithm exploration. Procedia computer science, 132, 880-889. https://www.sciencedirect.com/science/article/pii/S1877050918308329