Distributed agent-based air traffic flow management
Proceedings of the 6th international joint conference on Autonomous agents and multiagent systems
| Hi-index | 0.00 |
Evolutionary and genetic algorithms have been shown to be successful at optimizing air traffic flow, a critical component of the next generation air traffic system, designed to facilitate the future expansion of air traffic with little increase in infrastructure. However, while these methods work well with certain forms of genome structure they could potentially have difficulty scaling in more general search spaces on large air traffic problems. This abstract proposes a solution to this scaling problem where the full problem is decomposed into many components, where good component solutions can be learned or evolved individually. When done properly, the components can then be put back together to form a full solution. This method is tested on an air traffic flow problem, where a genome specifies air and ground holdings needed to keep a set of flights within departure, landing and sector capacity constraints. On experiments using the entire US air space with over 5,000 flights, this proposed method shows that it can generate a solution within 20% of optimal and is five times better than a basic evolutionary algorithm using the same genome.