dwa_planner vs. base_local_planner

edit

I'll try to answer your questions with a set of bullets below:

• The dwa_local_planner supports velocity constraints specified in x,y, and theta while the base_local_planner only supports constraints specified in x and theta. There is some support for y velocities, but users are limited to a pre-specified list of valid y velocity commands. This makes the dwa_local_planner a better choice for robots that are holonomic or pseudo-holonomic because it allows for better sampling of the velocity space.

• The dwa_local_planner is essentially a re-write of the base_local_planner's DWA (Dynamic Window Approach) option. The code is, in my opinion, a lot cleaner and easier to understand, particularly in the way that trajectories are simulated. I believe that much of the overlapping code between the two packages is factored out and shared. However, its true that there is some more work that could be done to pull out common functionality. For applications that use the DWA approach for local planning, the dwa_local_planner is probaly the best choice.

• Robots with low acceleration limits may wish to use Trajectory Rollout over DWA. This option is only available through the base_local_planner package. Documentation on the differences between Trajectory Rollout and DWA is available here.

• Both the base_local_planner and dwa_local_planner packages are actively maintained and stable though no new features are planned for either at the moment. As of the electric release, both planners can be tuned via dynamic_reconfigure.