1.10.3 (2013-03-08)

Removed PreferForwardCostFunction from dwa_local_planner, as it was causing problems when the initial direction was behind the robot. (Robot would hesitate to turn in place.)
Reintroduced capability lost in refactor of dwa_local_planner which allows intermediate failures in the map_grid_cost evaluation. It was preventing robot from turning in place during start of plan execution.

1.10.2 (2013-02-04)

TrajectoryPlanner: added deceleration to goal position.
TrajectoryPlannerROS: re-enforce min-rot-in-place-vel after acceleration limits are applied
AMCL: added request_nomotion_update service
Fixed rosbag invocation in some tests.
Fixed issue #18: The gradient-descent portion of navfn planner would sometimes fall into terminal oscillations before reaching the end of the path, and that has been fixed.
Removed references to common_rosdeps.
Fixed issue #16: path length limit increased.
AMCL: clear out laser/frame mapping when a new map is received.
Fixed bug #13 by setting acceleration limits higher.

Fixed bug introduced during dwa_local_planner refactor where Bresenham ray-tracing code used std::vector to return points inside the inner planning loop, which caused lots of malloc() calls which slowed it down to the point where the PR2 motion was jerky.
Fake_localization: added transform_tolerance param to match amcl.
!TrajectoryPlannerROS: added back checkTrajectory() and scoreTrajectory() functions removed in refactor. They were needed in navigation_experimental.