Switch branches/tags
Nothing to show
Commits on Dec 19, 2012
  1. Moved WITH-DESIGNATORS to cram_plan_library.

    moesenle committed Dec 19, 2012
    Additionally, made it more flexible:
     * A task tree node is now created.
     * A function is used for creating designators that uses hooks. This
       allows to, for instance, use an already instantiated designator, for
       instance produced by projection.
  2. Use the result of MONITOR-PROCESS-MODULE instead of PM-EXECUTE.

    moesenle committed Dec 19, 2012
    This should make high-level plans work with synchronous and asynchronous
    process modules.
Commits on Dec 18, 2012
  1. Got rid of roslisp-hotpatches.

    moesenle committed Dec 18, 2012
    The patches are now upstream in roslisp, so the hotpatch file is not
    required anymore.
Commits on Dec 14, 2012
  1. Revert "Removed call to monitor until we merge in the corresponding b…

    moesenle committed Dec 14, 2012
    This reverts commit 5298666.
  2. Moved reference call in AT-LOCATION to the parent task.

    moesenle committed Dec 14, 2012
    That way, it is possible to handle designator errors without entering
    the debugger.
  3. Fixed and improved AT-LOCATION.

    moesenle committed Dec 12, 2012
    It seems like AT-LOCATION still suffered from race conditions. Added a
    lock around anything related to navigation to prevent the reached
    predicate to succeed while the robot is navigating.
    Pulled out the at-location functionality into a function. That way, we
    do not need to recompile everything that uses the macro.
  4. GET-CACHED-COSTMAP needs to return the same costmap for all equated d…

    moesenle committed Dec 12, 2012
    Instead of regenerating a separate costmap for each equated designator,
    the function GET-CACHED-COSTMAP needs to return the same costmap. This
    is now solved by using FIRST-DESIG.
Commits on Dec 10, 2012
  1. Use LOCATION-DESIGNATOR-REACHED instead of perceive-state.

    moesenle committed Dec 10, 2012
    It seems like the function LOCATION-DESIGNATOR-REACHED behaves
    differently to perceive-state. Use it everywhere in AT-LOCATION to
    prevent location-lost errors.
  2. Added cram function NAVIGATE to allow for enforcing navigation.

    moesenle committed Dec 10, 2012
    When using (ACHIEVE `(LOC Robot ?loc)), the robot didn't move if it was
    already standing at a valid solution. This is not what we want in
  3. Fixed another bug in at-location.

    moesenle committed Dec 10, 2012
    Since TF callbacks are normally executed in a ROS subscriber thread, we
    cannot simply create a task object. Creating a top-level task doesn't
    work either since we need to inherit task-local variables. The new code
    now uses TV-CLOSURE in a thread to explicitly bind task local variables.
Commits on Dec 6, 2012
  1. Don't use a thread but a cpl:task for pulsing tf changes.

    moesenle committed Dec 6, 2012
    This ensures that all task-variable bindings are established correctly.
  2. Reindent.

    moesenle committed Dec 6, 2012
  3. Bugfix in at-location.

    moesenle committed Dec 6, 2012
    We need to dynamically rebind *tf* in the pulse thread that is executed
    at TF updates. Otherwise, the global tf listener, not a projection local
    one might be bound which causes a lot of errors.
Commits on Dec 5, 2012
  1. Added support for choosing different sampling functions.

    moesenle committed Dec 5, 2012
    The generation functions now have a SAMPLING-FUNCTION keyword parameter.
Commits on Nov 28, 2012
  1. Do not block TF in a long-running callback.

    moesenle committed Nov 28, 2012
    Pose verification was done in a TF callback. Unfortunately, it could be
    pretty expensive so it was screwing up TF. Instead of executing the
    recalculation in the TF thread, spawn a temporary thread if not already
Commits on Nov 27, 2012
  1. Fixed a bug in scaling of meshes.

    moesenle committed Nov 27, 2012
    Instead of outputting face instances, the code was just creating a
    vector of vertex-point lists.
Commits on Nov 26, 2012
  1. Refactored parsing of meshes.

    moesenle committed Nov 26, 2012
    Instead of just merging all mesh points, ignoring their position in the
    scene, we walt the tree of nodes now and generate the overall mesh using
    the transform in the nodes.
  2. Added aiNode to groveler.

    moesenle committed Nov 26, 2012
Commits on Nov 22, 2012
  1. LOAD-3D-MODEL can deal with multi-mesh files now.

    moesenle committed Nov 22, 2012
    Instead of just loading one mesh from an assimp file, the resulting mesh
    now contains all meshes made of triangles, merged to one single mesh to
    better support DAE files.