From: Mark M. <mm...@ri...> - 2014-07-28 18:52:41
|
Hi Pablo, We have something along those lines in the works. We expect to have an initial version out after the summer. On Jul 28, 2014, at 1:39 PM, Pablo Bustos <pb...@un...> wrote: > Thanks for the explanation Mark, > > but just to be clear, PRM-like planners cannot be restarted a second time with the previously learnt roadmap? and if so, "long-life" learning of a robot's environment cannot be performed, at least without serious work on the code? > > Thanks, > > Pablo > > > > > On Mon, Jul 28, 2014 at 6:36 PM, Mark Moll <mm...@ri...> wrote: > Hi Pablo, > > On Jul 28, 2014, at 10:30 AM, Pablo Bustos <pb...@un...> wrote: > > is it possible to save PRM's roadmap to disk after a running session and later initialize the planner with that same file? > > Any planner in OMPL can return the states and edges between them in the form of a PannerData object. Such objects can be stored and loaded again via the PannerDataStorage class. See ompl/demos/PlannerData.cpp for a demo program. The internal data structures of each planner are not preserved in PlannerData, so you can’t just restart a planner from previously computed PlannerData. Although that would be useful, it’s also a lot of work to serialize/deserialize the internal data structures for each planner. > > -- > Mark Moll > > > > > > > -- > Pablo Bustos García de Castro > Laboratorio de Robótica y Visión Artificial > http://robolab.unex.es > Escuela Politécnica > Universidad de Extremadura > Avda de la Universidad s/n > Cáceres 10003 > Phone: +34 927 257259 > Fax: +34 927 257203 > > > -- Mark Moll |