University of Nottingham > School of Computer Science > Automated Scheduling Planning and Optimisation > Air Transportation Research


Benchmarking of existing and new approaches for consistency and comparability

Since optimisation of airport operations is a real-world problem, it is important to consider both the real objectives of the airports and the real constraints upon the problems. The majority of the published work has considered real airport settings and it is apparent that both the objectives and the details of the constraints have differed between airports. Consequently, the models for the problems have also differed, resulting in the development of different solution approaches.

We think that it is desirable to have more consistency in this research area of airport operations. We wish to both encourage further research by those who may not have the direct contact with an airport and provide generic benchmark scenarios which can be used to quantify new and existing algorithms. It would also be of interest to evaluate the differences in quality and solution times between different exact (such as mixed integer linear programming) and heuristic (such as genetic algorithms) solution approaches.

For these reasons, we have set up a repository here for airport operations datasets. Further contributions from already published or new work are very welcome.

Ground Movement

Generic Ground Movement Problems

Layout Description Downloads
Dataset 1
Size : 12 nodes, 11 edges, 6 aircraft
Data input : static
Format : A description can be found here
This layout illustrates a relatively easily solvable scenario with one runway for arrivals (top left), one for departure (top right), 4 gates (bottom) and (relatively) independent taxi routes for arrivals and departures.
Dataset 2
Size : 10 nodes, 9 edges, 6 aircraft
Data input : static
Format : A description can be found here
This layout illustrates a harder scenario to solve, again with one runway for arrivals (top left), one for departures (top right) and 4 gates (bottom), but a shared taxi path between them.
Dataset 3
Size : 28 nodes, 33 edges, 8 aircraft
Data input : static
Format : A description can be found here
The details for this scenario can be found in (Roling and Visser, 2008)

Ground Movement@Manchester Airport

These datasets were generated from real data and are explained in the following paper, which was first presented at the PATAT 2014 conference in York:
Airport Ground Movement: Real World Data Sets and Approaches to Handling Uncertainty, Brownlee et al.
The data format is explained here.
You can obtain the source code for the tools to generate the layouts, which are mentioned in the paper
here (Ground movement tools wiki).

Benchmark Name Size Description Downloads
Manchester Benchmark 1 624 nodes, 666 edges, 3495 aircraft movements These are static instances representing real ground movement instances for Manchester Airport between the 29th of August and 3rd of September 2011. A description of the format can be found here
Manchester Airport Ground Movement Benchmark 2 624 nodes, 666 edges, real aircraft movements (503-1019 for 1 day, 2980-4676 for 6 days) These are duplicates of MAN Benchmark 1, with varying traffic levels. Flights are either randomly deleted or duplicated (with runway times corrected to keep 2 min separations); gates allocated with a stochastic method following operating rules at MAN. A description of the format can be found here
  • Benchmark Set
  • Paper detailing gate allocation method under review
  • Paper detailing data sets to be submitted

Airport Ground Movement layouts derived from OpenStreetMap

Ground movement layouts for several international airports, without aircraft movements. These datasets were generated from freely available mapping at OpenStreetMap. The data format and tools used to generate the layouts are the same as for the Manchester benchmarks above. Each layout is also supplemented by a KML file, allowing it to be visualised in Google Earth (TM). Note - these data sets are for non-commerical research purposes only. They are not intended for navigation.

Airport Size Description Downloads
Cologne-Bonn Airport (CGN) 1439 nodes, 1511 edges The sixth largest airport in Germany by aircraft movements. It has three runways and 110 stands excluding shadows (164 total).
Dubai Airport (DXB) 2249 nodes, 2666 edges A major international airport serving United Arab Emirates, one of the busiest in the world by passenger movements and more A380 movements than any other airport (21% of all A380 movements in May 2014). It serves as the hub for Emirates and FlyDubai, and a secondary hub for Qantas. It has two runways and 182 stands excluding shadows (185 total). The taxiway graph is densly connected compared with the other airports represented here, having parallel taxiways to access most stands.
Manchester Airport (MAN) 624 nodes, 666 edges The third largest airport in the UK by aircraft movement. It has two runways and 94 stands excluding shadows (148 total). There are several taxiways crossing the main runway, and a bottleneck where stand occupancy can block one taxiway, meaning that all northern stands are accessed by a signle taxiway
Melbourne Airport (MEL) 584 nodes, 631 edges The sixth largest airport in Australia by aircraft movements (although the second largest in terms of passenger numbers, several regional airports with high numbers of small aircraft have more movements). It serves as a major hub for Qantas and Virgin Australia. It has two runways and 50 stands excluding shadows (85 total).
Stuttgart Airport (STR) 980 nodes, 1058 edges The seventh largest airport in Germany by aircraft movements. It has one runway and 58 stands excluding shadows (63 total).
Sheremetyevo Airport (SVO) 3286 nodes, 3646 edges The second largest airport in Russia by aircraft movements, one of three major airports serving Moscow, and the hub airport for Aeroflot Russian Airlines. It has two runways and 181 stands excluding shadows (242 total).

Arlanda Airport

Airport Size Description Downloads
Stockholm Arlanda Airport (ARN) 624 nodes, 685 edges Arlanda airport is the largest airport in Sweden, the third largest airport in the Nordic countries and has 3 runways.

Reserve Crew Scheduling Data

Airport Size Description Downloads
Reserve Crew Scheduling Benchmark 2 days, 566 flights, 74 aircraft, 209 teams of crew, 12 reserve crew This test schedule is based on KLM operations during August 2012. The aircraft routings are those which were actually scheduled. The scheduled departure times have been modified to make the test instance more challenging in terms of the risk of delay propagation. To achieve this effect the departure times were set to the expected arrival time of the previous flight (on the same aircraft routing) plus the minimum turn time, for the given aircraft fleet type. The crew were scheduled using a set partitioning formulation solved in CPLEX.
  • The schedule and an explanation of different fields is provided in this zip archive.