Dynamic programming for aircraft routing using probabilistic global turbulence forecasts

A global turbulence nowcast and forecast system based on empirical data mining has been developed to provide gridded turbulence information for NextGen and the World Area Forecast System. Modeled on the FAA's Graphical Turbulence Guidance (GTG) product, the new system fuses model and observational data to provide 3-D, probabilistic, 0-36 hour turbulence nowcasts and forecasts globally, accounting for clear-air turbulence, mountain wave turbulence, and convectively-induced turbulence (CIT). This paper addresses the question of how to use these probabilistic forecast grids at various lead times to plan aircraft routes that balance the need to conserve fuel with the avoidance of potential hazards. The focus is on the problem of routing a single aircraft to a target airport over oceanic airspace given probabilistic hazard forecasts and deterministic upper-level wind forecasts. Although safe separation of multiple aircraft is not a considerable constraint in this scenario, extensions to the case of congested airspace is also discussed. The solution methodology is based on dynamic programming, casting the routing problem as a Markov decision problem and solving Bellman's equation iteratively.