Deep reinforcement learning for the olfactory search POMDP: a quantitative benchmark.

The olfactory search POMDP (partially observable Markov decision process) is a sequential decision-making problem designed to mimic the task faced by insects searching for a source of odor in turbulence, and its solutions have applications to sniffer robots. As exact solutions are out of reach, the challenge consists in finding the best possible approximate solutions while keeping the computational cost reasonable. We provide a quantitative benchmarking of a solver based on deep reinforcement learning against traditional POMDP approximate solvers.

Generation of momentum transport in weakly turbulent β-plane magnetohydrodynamics.

Magnetohydrodynamic turbulence on a β plane with an in-plane mean field, a system which serves as a simple model for the solar tachocline, is investigated analytically and computationally. We first derive two useful analytic constraints: We express the mean turbulent cross-helicity in terms of the mean turbulent magnetic energy, and then show that (for weak turbulence) the time-averaged momentum transport in the system can be expressed in terms of the cross-helicity spectrum. We then complete a closure of the system using weak turbulence theory, appropriately extended to a system with multiple interacting eigenmodes.

Turbulence model reduction by deep learning.

A central problem of turbulence theory is to produce a predictive model for turbulent fluxes. These have profound implications for virtually all aspects of the turbulence dynamics. In magnetic confinement devices, drift-wave turbulence produces anomalous fluxes via cross-correlations between fluctuations.

Ion-ion dynamic structure factor of warm dense mixtures.

The ion-ion dynamic structure factor of warm dense matter is determined using the recently developed pseudoatom molecular dynamics method [Starrett et al., Phys. Rev.