Robotics StackExchange | Archived questions

Incremental search algorithm in Ros_navigation_stack?

Hi! I want to use ros navigation stack alongwith gazebo for path planning simulation in a dynamic environment. I am stuck at 2 points and need some help-

  1. Ros navigation stack currently uses Dijkstra's algorithm or A* for global planner. Both these algorithms work for a static map. Is it possible to implement incremental search algorithms like D* or D* lite in global planner?

  2. How does navigation stack avoid dynamic obstacles? Does it call global planner and recompute the path to goal after adding those obstacles in the map?

Any help is highly appreciated.

Asked by sumitsk on 2016-04-02 23:19:35 UTC

Comments

Answers

  1. Yes, you can implement any planner you like by using the plugin interface.
  2. Implicitly yes. The local planner should try to avoid dynamic obstacles on its own, and if it can not, will return false. This should trigger the global planner again.

Asked by dornhege on 2016-04-03 08:39:18 UTC

Comments

I'm not sure D* will be as straightforward because it would probably require modifications to the internal move_base state machine and/or the planner interface. Related discussion: http://answers.ros.org/question/229221/is-it-possible-to-use-an-any-time-global-plannner-in-move_base/

Asked by spmaniato on 2016-04-03 15:57:03 UTC