Difference between revisions of "Project 2014"
| Line 1: | Line 1: | ||
{{FULLPAGENAME:Summer 2014 Research Plans}} | {{FULLPAGENAME:Summer 2014 Research Plans}} | ||
| + | |||
{{DISPLAYTITLE:<span style="display: none">{{FULLPAGENAME}}</span>} | {{DISPLAYTITLE:<span style="display: none">{{FULLPAGENAME}}</span>} | ||
Motor control often depends on accurate timing of motor commands, but animals succeed with amazing motor feats despite the obvious fact that biological timekeeping falls well short of perfection. While all known clocks keep time imperfectly, engineered clocks are good enough that their imperfections can usually be ignored. For this reason, perhaps it is not surprising that until very recently, nearly all formulations of control theory assumed that control agents had access to perfect clocks. With this classical control theory, it remained impossible to understand how animals, with substantial imperfections in internal timing, perform as well as they do. Over the last few years, my collaborators (at Johns Hopkins and elsewhere) and I have been working to extend control theory to apply when clocks remain substantially inaccurate. I contributed to this effort by publishing a paper [Carver, Fortune and Cowan, 2013] considering cooperative control of a two-person saw. We showed that lumberjacks could stay synchronized, despite substantially imperfect timing. In our model, a follower updated its clock by observing the saw and knowing (or successfully learning) the saw's nominal trajectory. Our demonstration required that we assume a statistical model for the internal clocks. In the absence of data to constrain this choice, we chose an inverse Gaussian process. More recently, preliminary work with data showed this simplest formulation to be a bad choice for parsimoniously describing reality [accomplished in collaboration with two Johns Hopkins students: Alexander Spinos, undergraduate, and Robert Nickl, Ph.D. candidate]. This summer (2014), I am looking to extend these results using biological time series to search for a better model of the internal clock. | Motor control often depends on accurate timing of motor commands, but animals succeed with amazing motor feats despite the obvious fact that biological timekeeping falls well short of perfection. While all known clocks keep time imperfectly, engineered clocks are good enough that their imperfections can usually be ignored. For this reason, perhaps it is not surprising that until very recently, nearly all formulations of control theory assumed that control agents had access to perfect clocks. With this classical control theory, it remained impossible to understand how animals, with substantial imperfections in internal timing, perform as well as they do. Over the last few years, my collaborators (at Johns Hopkins and elsewhere) and I have been working to extend control theory to apply when clocks remain substantially inaccurate. I contributed to this effort by publishing a paper [Carver, Fortune and Cowan, 2013] considering cooperative control of a two-person saw. We showed that lumberjacks could stay synchronized, despite substantially imperfect timing. In our model, a follower updated its clock by observing the saw and knowing (or successfully learning) the saw's nominal trajectory. Our demonstration required that we assume a statistical model for the internal clocks. In the absence of data to constrain this choice, we chose an inverse Gaussian process. More recently, preliminary work with data showed this simplest formulation to be a bad choice for parsimoniously describing reality [accomplished in collaboration with two Johns Hopkins students: Alexander Spinos, undergraduate, and Robert Nickl, Ph.D. candidate]. This summer (2014), I am looking to extend these results using biological time series to search for a better model of the internal clock. | ||
Revision as of 21:34, 30 December 2013
Summer 2014 Research Plans
{{DISPLAYTITLE:}
Motor control often depends on accurate timing of motor commands, but animals succeed with amazing motor feats despite the obvious fact that biological timekeeping falls well short of perfection. While all known clocks keep time imperfectly, engineered clocks are good enough that their imperfections can usually be ignored. For this reason, perhaps it is not surprising that until very recently, nearly all formulations of control theory assumed that control agents had access to perfect clocks. With this classical control theory, it remained impossible to understand how animals, with substantial imperfections in internal timing, perform as well as they do. Over the last few years, my collaborators (at Johns Hopkins and elsewhere) and I have been working to extend control theory to apply when clocks remain substantially inaccurate. I contributed to this effort by publishing a paper [Carver, Fortune and Cowan, 2013] considering cooperative control of a two-person saw. We showed that lumberjacks could stay synchronized, despite substantially imperfect timing. In our model, a follower updated its clock by observing the saw and knowing (or successfully learning) the saw's nominal trajectory. Our demonstration required that we assume a statistical model for the internal clocks. In the absence of data to constrain this choice, we chose an inverse Gaussian process. More recently, preliminary work with data showed this simplest formulation to be a bad choice for parsimoniously describing reality [accomplished in collaboration with two Johns Hopkins students: Alexander Spinos, undergraduate, and Robert Nickl, Ph.D. candidate]. This summer (2014), I am looking to extend these results using biological time series to search for a better model of the internal clock.
