The navfn code that you're referring to isn't actually code I wrote, though I do maintain the navigation stack. Its possible that the author will have more to say, but I don't have great insight in the best way to use/change navfn's A* implementation for D* Lite. Visualizing the potential array in rviz seems like a valuable thing to do, however, if you want to submit a patch for that, I'd be happy to take it back to navigation.

Eitan

On Tue, Dec 21, 2010 at 11:28 AM, Ye Cheng <cnhzcy14@gmail.com> wrote:

Hi Eitan,

Here is the story. I have a need of Incremental heuristic search in Base Global Planner for long range navigation in an unknown area, so I tried to modify your A* in move_base to D* Lite. However, it is really difficult to arrange the order of the re-explored cells in the priority buffer because you are using the threshold curT to control the behavior of the priority queue. And also, some re-explored cells have to be updated twice by the D* Lite during the re-planning. So I take use of your code to realize a new way: 1) Keep the potarr during one long range move; 2) Swap the start and goal cells before the propagation, make your A* as a backward version; 3)Do a limited re-propagation when re-planning is necessary. I also create a PointCloud2 message for the explored potential area, and add this patch to visualize the message. Here is a screencast of my experiment.

At the moment, this method is not stable yet, some times the robot will be stopped by the incomplete potential area. Do you think it is a good way to do the Incremental replanning? Or it is still possible to realize the D* Lite based on your A*？

Cheers,
Simon

_______________________________________________
ros-users mailing list
ros-users@code.ros.org
https://code.ros.org/mailman/listinfo/ros-users