BEGIN:VCALENDAR VERSION:2.0 PRODID:-//talks.cam.ac.uk//v3//EN BEGIN:VTIMEZONE TZID:Europe/London BEGIN:DAYLIGHT TZOFFSETFROM:+0000 TZOFFSETTO:+0100 TZNAME:BST DTSTART:19700329T010000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0100 TZOFFSETTO:+0000 TZNAME:GMT DTSTART:19701025T020000 RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT CATEGORIES:Isaac Newton Institute Seminar Series SUMMARY:Using Data-driven Methods to Understand and Contro l Active Materials - Michael Hagan (Brandeis Unive rsity) DTSTART;TZID=Europe/London:20250910T143000 DTEND;TZID=Europe/London:20250910T151000 UID:TALK233293AThttp://talks.cam.ac.uk URL:http://talks.cam.ac.uk/talk/index/233293 DESCRIPTION:Active matter is composed of particles that genera te forces or motion\, which leads to spectacular e mergent dynamics. In principle\, active matter cou ld form the basis for a new class of materials wit h lifelike properties of biological organisms. Yet \, active materials exhibit diverse dynamical stat es\, most of which have chaotic dynamics that do n ot produce work or other useful functions. Thus\, a robust control strategy is needed to drive activ e materials into an emergent state that correspond s to a desirable function. However\, designing con trol protocols requires accurate dynamical models\ , which are not available for most active matter s ystems. Developing quantitative models using tradi tional statistical physics approaches is challengi ng because active materials lack the scale separat ion characteristic of equilibrium systems. \;I n this presentation\, I will discuss efforts to co mbine machine learning\, other data-driven techniq ues\, and  \;physics-based models with control theory to address this challenge in the context o f a widely-used active material\, microtubule-base d active nematics. Recent advances in optogenetic motors have enabled constructing the light-activat ed active nematics\, in which the activity can be spatiotemporally controlled by shining light on th e sample. The challenge is to determine the spatio temporal light sequence required to drive the syst em into a desired behavior.I will describe two com plementary approaches to computationally determine an optimal light sequence. In the first\, we have adapted a method to discover optimal physics-base d continuum models directly from spatiotemporal da ta\, using sparse regression. We have identified s everal approaches to mitigate measurement errors i n the data. We find that the method can reveal the relative contributions of different physical mech anisms\, and quantitatively estimates key experime ntal parameters. Then\, we have developed a framew ork to combine the optimal physics-based model wit h optimal control theory to solve for the spatiote mporal activity profile that drives the system int o a desired state. We demonstrate that active mate rials can be driven into arbitrary behaviors\, inc luding those which do not correspond to dynamical attractors and thus cannot be accessed without con trol.Since no model is perfectly accurate for a sp ecific system\, in the second approach we develop a deep reinforcement learning (DRL) based controll er to enable model-free control of active material s. The controller discovers and implements spatiot emporal sequences of activity to drive a 2D active nematic system toward a prescribed dynamical stea dy-state. This framework does not require a detail ed physics model\, making it ideal for complex act ive materials that lack quantitative theoretical d escriptions. Furthermore\, the approach is extreme ly robust to noise and experimental measurement er ror. We compare the performance of the physics-bas ed and DRL-based controllers for active nematics.T his work was supported by DE-SC0022291. Preliminar y work was supported by NSF DMR-1855914 and DMR-20 11846. Computing resources were provided by XSEDE TG-MCB090163 and the Brandeis HPCC (DMR-MRSEC 2011 846 and OAC-1920147). LOCATION:External CONTACT: END:VEVENT END:VCALENDAR