Airborne self-separation at innovation workshop

The iFly project

In recent years the trend has been to direct airborne self-separation research projects to situations of less dense airspace. This trend is surprising because airborne self-separation was originally conceived as a potential solution for high density airspace. Now, however, two schools of researchers exist. Researchers of one school believe that airborne self-separation can safely accommodate traffic levels much greater than current en-route traffic. Researchers of the other school believe that airborne self-separation cannot work safely for high density airspace. Both schools concur that for airspace having sufficiently low traffic densities, airborne self-separation may be safe. From a research perspective this calls for a study which evaluates up to which traffic levels airborne self-separation is safe. This is exactly the key aim of the European Commission project iFly.

The iFly project began in May 2007 and will end in August 2011. There are 18 members of the consortium, the majority of which are universities, complemented by research organisations and including an air navigation service provider and an aviation equipment manufacturer. One of the first iFly project deliverables is an initial design of an advanced airborne self separation concept of operations, referred to as the Autonomous Aircraft Advanced (A3) ConOps. At the same time, various specific iFly studies are continuing in order to develop further improvements and refinements of the initial A3 ConOps.
  HTML iFly website
Henk Blom
Project coordinator  - NLR

Studies presented at the innovation workshop

At the innovation workshop, held at the EUROCONTROL Experimental Centre in December 2009, results obtained from three such iFly studies were presented.
  • Comparison of self-separation priority rules
  • Analysis of complex, multi-agent hybrid systems
  • Mid- and short-term conflict resolution in autonomous aircraft operations

Comparison of self-separation priority rules

In common with some earlier self-separation concepts, the initial A3 concept proposes the use of pair-wise priority rules to designate an aircraft which must modify its trajectory in order to resolve a predicted proximity with another aircraft. In order to better understand and possibly improve the initial A3 concept, a conflict resolution algorithm has been used to simulate self-separation using priority rules and this has illustrated situations in very dense traffic in which a designated aircraft is not able to resolve a predicted proximity.
In the above example, all aircraft are moving vertically. AEA0520 (blue) is predicted to lose separation with AF40063 (pink, bottom left). AEA0520 (blue) is designated to resolve the conflict. It cannot do so because it is boxed in to the left by RAM0497, to the right by BAW1183 and below by AEA0498. On the other hand, AF40063 could resolve the conflict.

A resolution strategy is less likely to fail if the designation of the aircraft responsible for manoeuvring can take into account the feasibility of finding a conflict-free trajectory for that aircraft. One simple way of doing this is to allow priorities to be reversed if the first designated aircraft is unable to find a conflict-free trajectory. Further simulations allowing priority reversal only revealed irresolvable situations with short air-air datalink ranges.
  Acrobat Paper, "Comparison of pair-wise priority-based resolution schemes through fast-time simulation"
  Acrobat Presentation
Richard Irvine
EUROCONTROL Experimental Centre

Analysis of complex, multi-agent hybrid systems

Verifying the correct behaviour of complex air traffic management systems is difficult. Given the demanding safety requirements, formal methods that can guarantee correctness are very appealing, but the complexity of the problem makes their application challenging. Reasons for these difficulties include:
  • Heterogeneity: agents operating in ATM systems such as aircraft, electrical devices, and human beings, are characterized by fundamentally different mathematical models so that their interaction is difficult to analyze and model.
  • Size of the system: the number of agents in realistic ATM scenarios is generally very large.
The aim is to benefit from using a hybrid systems formalism to describe the behaviour of ATM systems, both in normal and abnormal operating conditions. Indeed, it is possible that situations, which are not critical when considering agents in isolation, become critical when considering their interaction. In this paper, the authors consider a formal notion of composition which models the interaction of the agents involved in ATM systems. By applying this notion of composition to agents, a mathematical model is developed that can be used to formally analyse and detect the occurrence of unsafe and/or not allowed situations in the overall ATM system. Since the number of agents operating in realistic ATM scenario is large, the mathematical model is difficult to analyse. To cope with this problem, a complexity reduction approach is presented that is based on the decomposition of the set of critical situations of the overall ATM system into smaller sets, whose detection can be performed with reasonable computational effort. In particular, this method allows the analysis of ATM scenarios with an arbitraryily large number of agents. As an example, the proposed methodology is applied to the ASAS (Airborne Separation Assistance System) Lateral Crossing Procedure.

  Acrobat Paper, "A Compositional Hybrid System Approach to the Analysis of Air Traffic Management Systems"
  Acrobat Presentation
Giordano Pola
University of L'Aquila
Maria Domenica Di Benedetto
University of L'Aquila

Mid- and short-term conflict resolution in autonomous aircraft operations

The conflict resolution problem was investigated in a self-separation airspace, in accordance with the iFly A3 concept of operations. A hierarchical control scheme has been developed, which makes use of two techniques: navigation functions are used to resolve conflicts arising in the short term and model predictive control acts in the mid term.

The idea behind navigation functions is straightforward: each aircraft can be viewed as a positive charge in a potential field, while its destination is a negative charge. Thus, each aircraft is attracted towards its destination and repelled by all other aircraft. This method has been proved to provide conflict free trajectories, provided that initially the aircraft are not already in conflict and that aircraft can violate airspeed and turn rate constraints. The technique was originally developed for robots, which are not subject to these kinds of constraints (they can stop, turn around, etc.) - constraint violation has not been tackled within this method.

Model predictive control is a technique specifically designed to handle constraints. It is used in the mid term to force aircraft performance constraints (airspeed and turn rate) on the navigation functions. The combined control scheme can optimise aircraft trajectories, maintaining the guaranteed conflict avoidance that navigation functions provide, while respecting aircraft constraints. However, not all situations are resolvable anymore, as several degrees of freedom are lost by meeting the aircraft performance constraints. Nevertheless, the scheme is able to cope with typical traffic situations and is now being tested against higher density scenarios to determine its limits.

Figure 1 Hierarchical Conflict Resolution scheme
As shown in the above example of three aircraft, both the short and mid term algorithms operate in an autonomous fashion. The only communication needed is at the mid term level, where every aircraft should have information about the trajectory planned by the mid term algorithm of all other aircraft that are involved in a conflicting situation. In the short term, navigation functions coordinate implicitly (by construction) and thus, the only information they need is the current position of other aircraft.

The proposed scheme has been tested using artificial traffic scenarios, involving up to 8 aircraft, all converging on a point and the results are promising. The control scheme allows priorities to be considered, either constraining the aircraft with higher priority not to manoeuvre or indirectly penalising resolution advisories that require higher priority aircraft to alter their trajectories. Figure 2 shows an example for the case of 4 aircraft, where priorities are not considered.

Figure 2 Trajectories for 4 aircraft initially converging on a point, after conflict resolution
Currently the scheme is being evaluated using realistic traffic data. Furthermore, ways of considering human factors in the simulations are under investigation, in order to make the algorithm produce trajectories that are simpler for pilots to fly.
  Acrobat Paper, "Mid and Short Term Conflict Resolution in Autonomous Aircraft Operations"
  Acrobat Presentation
Georgios Chaloulos
Eidgenössische Technische Hochschule, Zürich
John Lygeros
Eidgenössische Technische Hochschule, Zürich
Giannis Roussos
National Technical University of Athens
Kostas Kyriakopoulos
National Technical University of Athens
  Last validation: 06/04/2010