Glider flight environment modeling for optimal control

This paper describes the process of creating a computationally-inexpensive yet relatively accurate atmospheric environment model for use in stochastic optimal control problems for glider flight management. In such problems, estimates of transition probabilities between flight in updrafts, downdrafts and thermals of varying strength are needed. This work proposes an atmospheric environment model that predicts updraft and downdraft strengths in a given region and, when combined with existing glider flight data, estimates thermal locations and strengths. The resultant predictions can be utilized to compute the desired transition probabilities. A simple approach currently employed in flight simulator games is adapted for updraft and downdraft modeling. The method is empirical and requires the computation of a linear factor. Interestingly, when validated against actual flight data, this technique is 92.4 percent accurate on average. This paper also shows that the location and intensity of thermals can be deduced from flight data by utilizing updraft and downdraft predictions. The work then illustrates the modeling process for a sample topographical area, and utilizes the model to solve a stochastic drift counteraction optimal control problem where control policies that maximize glider flight range are generated.