55-1446_SOW.docx

N00173-17-Q-0167

Statement of Work Oregon State University

Task 1: Oregon State University (OSU) will develop stochastic trajectory optimization algorithms for teams of underactuated underwater drifters. The developed techniques will build on Hollinger’s prior work in generalizing the STOMP algorithm to minimizing energy usage in cases where there are disturbances (e.g., ocean currents) affecting the vehicle’s motion. The developed algorithms will be compared to the MDP approaches developed at NRL in simulations on the SCS dataset to determine which algorithm provides superior performance in terms of: coverage, energy usage, and computational efficiency. The OSU team will also explore hybrid algorithms that combine the benefits of MDPs and stochastic trajectory optimizers.

Task 2: OSU will also provide processed data sets of ocean current data from the SCS dataset, along with estimated uncertainty measures on these quantities. The uncertainty measures will be generated using interpolation variance, which was shown to provide strong correlation with true errors in Hollinger’s prior work. OSU will also develop a method for updating ocean current models based on the executed trajectories of ocean drifters. These updates will take advantage of the predicted and true surfacing locations to adjust the model of ocean currents in that area. These updated model swill be incorporated into the stochastic trajectory optimization algorithms in Task 1.