Planner using : Smac Hybrid A* controller: DWB, Have to test with MPPI

I have two questions.

First Query

  1. I am trying to setup the NAV2 stack for Ackermann robot. However, the issue I am still facing is global planner struggling to generate plan as soon as I give a goal in rviz. - Terminal console : Cannot generate feasable plan, starting point in lethal space

I have enabled footprint clearing but no progress. As suggested in the issue here: https://github.com/ros-planning/navigation2/issues/3992.

My NAV2 params are same as mentioned in my above github issue.

I decided to try with simulated lidar to verify the problem, it seems still having the same issue. I still can't figure out which configuration should I change. This is the output. If I move the robot for some distance then it starts to generate the plan. But this scenario fails in the real-physical robot.

Another Thing, why the generated plan in the below image seems too complex? I observe these kind of plans many times.

enter image description here

Video link: https://drive.google.com/file/d/1FoM158Fm8Xsgw0Zv4nFmKKxhAdnBTFDj/view?usp=sharing

Also, after some more researching, I ma trying to launch the NAV2 with large empty map with free space only to make the planner generate the plan as mentioned here: GPS+Navigation2 : Issues navigating with an empty map after success with a pre-built map: "Received plan with zero length"

I have opened queries previously as well but hard luck. Also individual configuration advices are difficult to suggest but I think I am pretty close now to make the whole stack working. If any body could suggest some necessary changes, it will be really appreciated. @StevenMackenski

Second Query

I am working on an application for which I want the global planner (Smac Hybrid A*) to generate a plan from (say point A to B) for once but the don't update every time. How can I achieve this behaviour.

  • $\begingroup$ For the second question it's a duplicate refer to robotics.stackexchange.com/questions/104234/… $\endgroup$
    – akchobby
    Commented Jan 8 at 9:32
  • $\begingroup$ Do you have traversing unknown space as valid in your global costmap configurations? $\endgroup$ Commented Jan 8 at 19:49
  • $\begingroup$ I have enable track unknown pace as true in global costmap params. It seems the issue is initially the cell where the robot is getting considered as occupied, meaning footprint clearing is the issue. However, I have set the clearing param to true. $\endgroup$ Commented Jan 8 at 21:15
  • $\begingroup$ Also observed that robot base link is out of the map but for global planner it should be inside the map to generate the plan. I am having RGB-D sensor mounted in front, so how can the whole robot be in the map initially, if it can only see the front. That's why I am trying to create an large empty map free space which will just trigger the planner to generate plan irrespective of occupied or unknown space. @StevenMackenski The whole application for now is to just setup the general NAV2 stack for the robot with SLAM not with pre-built map. $\endgroup$ Commented Jan 8 at 21:23
  • $\begingroup$ Does this update seems similar to what I am facing. github.com/ros-planning/navigation2/pull/3980 $\endgroup$ Commented Jan 8 at 21:30


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.