This page lists changes that were made in the 1.3.x series of navigation.
For future plans, see the roadmap. See also navigation/Releases.
1.4.0 (2010-02-09)
The navigation stack now supports the sensor_msgs/PointCloud2 message which should allow it to be used directly with many PCL filters and nodelets. As a result of this, sensor_msgs/PointCloud support becomes slightly less efficient because a conversion must take place.
- Added the "map_topic" parameter which allows users to change the topic the costmap uses to subscribe to the map with. This allows for costmaps instantiated within the same node to subscribe to different topics.
Uses pcl types internally for storage of sensor_msgs/PointCloud2 structures rather than using sensor_msgs/PointCloud messages.
Moved the dwa_local_planner from the navigation_experimental stack to the navigation stack. This planner was used by the PR2 during its continuous_ops run and provides nice features like dynamically reconfigurable parameters. The package should have up-to-date documentation on the wiki explaining how things work. Though, its interfaces are very similar to the base_local_planner.
- Added Eric Perko's cost visualization tool to the base_local_planner package. This allows for visual feedback when tuning the cost function used by the base_local_planner.
- Now correctly searches upward for the "controller_frequency" parameter which is assumed to be the frequency at which the local planner is running. This actually removes a hard-coded value which should have disappeared a long time ago.
- Adding the latch_xy_goal_tolerance parameter to allow for considering the goal position to be reached permanently once it is ever achieved. This should help robots with poor odometry.
- Added an angular_sim_granularity parameter that allows setting the minimum angular simulation step independently from the translational simulation step.
Deprecated the backup_vel parameter in favor of escape_vel since the old name was misleading.
Added the meter_scoring parameter to allow users to set their goal and path distance bias factors independent of the resolution of the costmap.
- Added a scoreTrajectory function from Travis Deyle's patch to complement checkTrajectory, but return a true score.
Added the visualize_potential parameter that, along with Ye Cheng's patch, allows the potential area that navfn computes to be visualized in rviz.
Added the clear_costmaps service that allows an external user to tell move_base to clear all the information in its costmaps. Patch from JSK (manabusaito).
Moved the move_slow_and_clear recovery behavior from navigation_experimental to navigation.
1.3.1 (2010-11-24)
- Now executes recovery behaviors when the local planner fails as well as the global planner.
Executes recovery behaviors when the robot oscillates for too long. Added two new parameters to move_base.
- Patch from Andreas Tropschug for supporting different map topics for different costmaps within the same node. Adds the "map_topic" parameter to the costmap.
1.3.0 (2010-10-05)
- Fix from Eric Perko for incorrectly computing the maximum velocity to stop before getting too close to an object.
- Moving test to use rosbag instead of rosplay
Added support for changing global frames. This means that an OccupancyGrid can be given to the costmap in a frame that is different from the current global frame and things will still work.
Added warnings when an OccupancyGrid with an origin that specifies a rotation is passed to the costmap.
- Now executes recovery behaviors when the controller fails in addition to the global planner.
- Now handles the case where the global frame of the planner changes at runtime.
- Planner now supports a tolerance when making a plan so that if people want to use it to get to a general region, it still work. Say you want to go all the way across the building, but you don't care if you're two feet away from the actual goal point.