Abstract:
World air transport has been on a steady exponential rise since the 1940’s and the
trend has shown remarkable resilience to external shocks. The level of air traffic has
greatly exceeded the wildest expectations of the air traffic management pioneers that
originally defined the basic precepts ATM that persist till today.
This has stretched ATM to a point where it is starting to show signs of
ineffectiveness in the face of ever increasing congestion. Delays are on the rise, costs
are ballooning, flights are being elongated unnecessarily, the system is becoming
increasingly susceptible to disruption, and the high environmental impact of aviation is
being compounded by the inability of air traffic controllers to optimise ATM operation
in real-time. If these trends are not reversed, ATM could eventually face instability.
The conservative, self-preserving outlook of the ATM community has confined
progress to relatively minor tweaks of a tired human-centric paradigm. However, the
diverging gap between ATM performance and fundamental requirements indicates the
need for a step change.
In this work, the traditionally incremental approach to ATM research was broken
to favour a more exploratory mindset. As a result, a new discipline called
Computational Air Traffic Management has been defined to address the unique set of
challenges presented by the ATM problem, by taking a more objective scientific
approach.
A specific embodiment of a CATM system was designed, constructed, simulated
and tested and shown to be a significant step towards demonstrating the feasibility of a
fully autonomous multi-agent-based air transportation system based on optimisation
principles. The system offers unique advantages in terms of resilience to disruption,
efficiency and future scalability. The traffic density using such a system can be
realistically increased many times higher than current levels while significantly
improving on the current levels of safety, operating cost, environmental impact and
flight delays. This work advances the field of ATM as well as the fields of
Computational Intelligence and Dynamic Optimisation of High Dimensionality Non-
Convex Search Spaces.