A reinforcement learning approach for mixing in stratified shear flows

The prediction of transport in the ocean is limited by large uncertainties. In fluid flows, reinforcement learning has been used to discover navigation strategies, achieve flow control, and recover model parameters. In this work, we use an actor–critic deep reinforcement learning approach for optimal mixing in a stratified shear flow, a canonical model for flows in the ocean. In this continuous control setting, a reinforcement learning agent introduces small perturbations to the flow, with the goal of efficiently achieving a desired turbulent flux coefficient. We describe the reinforcement learning setup and present initial results on mixing in the stratified shear flow environment, where a learned strategy produces four times higher levels of buoyancy dissipation than a random strategy at the same level of energetic forcing.