【Lecture】Smart living and artificial particles to navigate in complex flows via Reinforcement Learning

Theme:  Smart living and artificial particles to navigate in complex flows via Reinforcement Learning

Time:  Monday, April 2, 2018 14:30-16:30

Venue: RM A348, Aerospace Building

Lecturer:  Professor Luca Biferale, Dept. of Physics, University of Rome, Italy

Bio:

Apr. 2014 - present Full Professor of Theoretical Physics, Mathematical and numerical modelling Dept. Physics, University of Rome “Tor Vergata” (Italy).

2008 Elected Fellow. APS, division of “Statistical and Nonlinear Physics”

2010 Elected Fellow. EUROMECH Society, division of “Fluid Dynamics”

Key numbers (scientific impact)

Number of published papers: 220+

Hirsch-index: H = 45 (Google Scholar)

i10-index = # publications with more than 10 citations: 120+ (Google Scholar) Citations: 6800+ (Google Scholar)

webpage: http://www.fisica.uniroma2.it/~biferale/

Abstract：

We present a numerical study to train smart inertial particles to target specific flow regions with high vorticity through the use of Reinforcement Learning (RL) algorithms (S. Colabrese, K. Gustavsson, A. Celani, and L. Biferale. arXive:1711.05853). The particles are able to actively change their size to modify their inertia and density. Using local measurements of the flow vorticity, the smart particle explores the interplay between its choices of size and its dynamical behaviour in the flow environment. This allows to accumulate experience and learn approximately optimal strategies of how to modulate the size in order to reach the target with high-vorticity regions. We consider flows with different complexities: a two-dimensional stationary Taylor-Green like configuration, a two-dimensional time-dependent flow, and finally a three-dimensional flow given by the stationary Arnold-Beltrami-Childress helical flow. We show that smart particles are able to learn how to reach extremely intense vortical structures in all the tackled cases. This study completes a series of previous explorative applications of RL for flow navigation [K. Gustavsson, L. Biferale, A. Celani, S. Colabrese, Eur. Phys. J. E 40, 110 (2017); S. Colabrese, K. Gustavsson, A. Celani, L. Biferale, Phys. Rev. Lett. 118, 158004 (2017)]