OASIS (Optimal Aircraft Sequencing using Intelligent Scheduling)

OASIS is a prototype air-traffic management system developed for Sydney's Kingsford Smith airport. OASIS accurately calculates estimated landing times, determines the sequence of aircraft to land giving the least total delay, and advises air traffic controllers of appropriate control actions to achieve this sequence. It also monitors and compares actual progress of aircraft against the established sequence, and notifies the air traffic controller of significant differences and appropriate action to correct the situation. OASIS is designed to be responsive to sudden changes in environmental conditions (such as meteorological conditions or runway configuration) and changes in user objectives (such as aircraft operational emergencies or requirements).

OASIS combines artificial intelligence, software agents, and conventional software techniques to achieve a versatility significantly superior to that of any other system designed as an aid to air traffic flow management.

Air Traffic Congestion and Airport Capacity

Air traffic congestion is now a major problem for air travellers, operators, aviation authorities and governments. A study on European air traffic congestion, prepared by SRI International for the International Air Transport Association (IATA) in April 1990, illustrates the magnitude of the problem.

The SRI study forecasts that air traffic congestion and the resulting constraints on growth will cause annual losses of US$10 billion to European national economies by the year 2000. SRI considered that sufficient airspace capacity exists to accommodate the expected growth until 2010, if it is organised efficiently. However, traffic through at least 10 major European airports will be limited by insufficient capacity between 1995 and 2000 unless their capacity is enhanced.

The traditional solutions to airport congestion problems have been to:

  • Build additional airports or other facilities (the capacity enhancement approach).
  • Restrict aircraft movements (the demand management approach).
  • Capacity enhancement has the disadvantage of high capital cost, while demand management restricts the scheduling flexibility of the airlines.

In comparison, OASIS aims to provide significant capacity gains within the constraints set by the existing infrastructure at relatively low cost.

The Australian Response

With air traffic congestion as a real and growing problem at a number of Australia's airports, the Civil Aviation Authority (now Airservices Australia) instituted a program to improve air traffic management. After reviewing current systems, the Authority decided that a system based on Artificial Intelligence (AI) offered the best prospects, and that the Australian Artificial Intelligence Institute was best placed to develop such a system with the Authority's assistance. The Australian Department of Industry Technology and Commerce was similarly impressed with the OASIS joint venture proposal and provided grants to underpin some development costs.

Flow Control

The capacity of an airport is determined by both the physical characteristics of the aerodrome and the prevailing weather conditions. Congestion and delays occur when the rate of arrival of aircraft exceeds this capacity.

Air Traffic Control Flow Managers are responsible for sequencing arriving aircraft and departures for the most efficient use of the available capacity. Working in a continuously changing environment the Flow Manager must: detect approaching aircraft on radar, estimate times of arrival, compute the most efficient sequence, assess the effect of wind, monitor the sequence, and determine control actions.

OASIS assists the Flow Manager by accurately calculating ETA's determining if the demand exceeds the actual capacity, computing the sequence, and then giving appropriate instructions to achieve this.

This gives the Flow Manager more time to concentrate on evaluating alternative control actions. They can manually reserve slots in the sequence and, if necessary, issue instructions other than those recommended by OASIS.

In this way, the system improves runway utilisation and air traffic management, avoiding the inflexibility that accompanies fully automated systems.

System Design

To possess these capabilities, the system must not only be able to create and execute plans, but also must be willing to interrupt or abandon a plan when circumstances demand it. Moreover, because the ATC environment is continually changing and other agents and processes can issue demands at arbitrary times, performance of these tasks requires an architecture that is both highly reactive and goal-directed.

Intelligent software agents have been utilized to achieve this performance:

  • Procedural reasoning techniques are used for performing the flow control procedures.
  • Real-time task management techniques are used to integrate the reactive and goal-directed tasks under stringent real-time constraints.
  • Heuristic search techniques are used to determine the optimum sequence.

Other numerically intensive tasks, such as ETA calculation, are performed using conventional software techniques.

The software agent approach to air traffic management provides a system with a self-monitoring capability able to adapt it's behaviour as required according to prior performance. The real-time kernel of OASIS, unlike conventional expert systems does not assume a static environment during the reasoning process. It is capable of reacting to changes in weather and aircraft behaviour as soon as they are observed.

The system possesses multiple process capability, including interprocess communications facilities. In OASIS, there are two main classes of software agents: those handling global or inter-aircraft co-ordination, and those performing local computation or reasoning for each aircraft.

In essence, there is one process dedicated to each aircraft known to the system, and a fixed, small number of global agents responsible for co-ordination and mediation between the various aircraft. These global agents perform the sequence optimisation and manage wind information and user interaction.

Features of OASIS

OASIS offers a number of features not found in currently operational systems:

  • It is based upon AI technologies, possessing a combination of goal-directed and reactive (data-directed) behaviour that is suited to the ATC environment.
  • The sequence can be optimised with respect to time delay only, or a combination of delay and nominated operational costs.
  • It sequences to multiple runways simultaneously, including runways that handle both departures and arrivals.
  • The sequence calculation can include allowance for variable spacing between aircraft for greater efficiency.
  • It can accommodate "pop up" aircraft (i.e. unexpected arrivals from airports within the sequence area) into the sequence.
  • The sequence is continuously monitored by the system, which suggests further actions where necessary.
  • The range at which sequence calculation can commence is limited only by the availability of radar data.
  • The system monitors its own performance using an internal wind model which is updated on the basis of observed aircraft behaviour. This allows the system to account for the current wind effect on ETAs.
  • The sequence can be re-calculated rapidly to accommodate runway changes.
  • The knowledge base is separated from the reasoning system. Procedural knowledge is entered via a graphical user interface.
  • Data such as the airport configuration or aircraft performance is held in a separate database.
  • Aircraft performance calculation is flexible, allowing incorporation of alternative aircraft performance models.
  • The system has been developed on a Unix workstation with X-windows, allowing it to be easily integrated with the new generation of distributed ATC systems.

The Future

OASIS can handle in excess of one hundred aircraft agents, each interacting with a number of scheduling and coordination agents. The prototype system successfully completed operational tests at Sydney airport in April 1995.

AAII is currently developing a commercial version of the Sydney-based prototype system, called HORIZON that will be suitable for installation at any airport. The initial release of this product is expected in late 1997.