From: Boris L. <la...@in...> - 2011-03-24 10:35:09
|
Hey folks, I'm new to ompl, so bare with me. Here's what I do: planning for a rectangular wheeled robot in x,y,theta. Just geometric path planning, no dynamics. The collision checks are done by lookups in distance maps generated after convolution with the robot shape. When I use the RRT, it returns a decent solution that is free of obstacles. However, when I try to use KPIECE or the default, it says "Found solution" and returns a path that looks completely insane and does not care about obstacles at all. My code is below... Could this be a problem with the projection missing, i.e., is the default not good enough? Do I have to initialize the KPIECE grid somehow? Do I need motion validity checking instead of state validity? And a last question, which planner do you recommend as a benchmark for comparison against my own voronoi-based gridmap planner? Any help is appreciated. Best, Boris ob::StateManifoldPtr manifold(new ob::SE2StateManifold()); manifold->as<ob::SE2StateManifold>()->setBounds(...); og::SimpleSetup ss(manifold); ompl::base::SpaceInformationPtr si = ss.getSpaceInformation(); si->setStateValidityChecker(boost::bind(&isStateValid, _1)); si->setStateValidityCheckingResolution(1.0/manifold->getMaximumExtent()); si->setup(); ss.setPlanner((const ompl::base::PlannerPtr)new og::RRT(si)); ompl::base::ScopedState<ompl::base::SE2StateManifold> start(manifold); ompl::base::ScopedState<ompl::base::SE2StateManifold> goal(manifold); [then setting valid values for start goal] ss.setStartAndGoalStates(ompl_start, ompl_goal); bool solved = ss.solve(10000.0); -- Boris Lau Albert-Ludwigs-University Freiburg Institute of Computer Science, Autonomous Intelligent Systems Group Georges-Koehler-Allee 079, D-79110 Freiburg, Germany Building 079, Room 1005 Phone: +49 761 203-8012 | Mobile: +49 174 9436758 http://www.informatik.uni-freiburg.de/~lau |