You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have been working with manipulating objects and noticed that often, we use a different pipeline for pose estimation of target objects (could be an AI model, or some other classical approach etc) and therefore have a collection of candidate pick/target poses for which we then try to plan a path for any of those target poses.
I was wondering if we had a generator stage where we could feed in a series of pre-computed target poses. I did not find anything like that but I did see there is GenerateGraspPose stage which takes in a single target pose.
Using that template, I was able to easily create a custom stage which instead takes in a Pose array and generates solutions for all those poses as a starting point for planning.
I feel this is a common requirement for a lot of manipulation pipelines, and was wondering if we could have that as an added stage in the moveit core itself!
(FYI: I use the humble branch)
The text was updated successfully, but these errors were encountered:
A PR extending the functionality of the GeneratePose stage is welcome. Did you extend the stage GenerateGraspPose or GeneratePose? The former only considers a single object frame and generates grasps from multiple approach directions, while the latter actually considers manually specified poses.
My mistake - I meant to say : I extended the GeneratePose stage to take an array of target poses instead of just a single one, thus allowing users to now compute a series of candidate poses to the generator.
Hello!
I have been working with manipulating objects and noticed that often, we use a different pipeline for pose estimation of target objects (could be an AI model, or some other classical approach etc) and therefore have a collection of candidate pick/target poses for which we then try to plan a path for any of those target poses.
I was wondering if we had a generator stage where we could feed in a series of pre-computed target poses. I did not find anything like that but I did see there is GenerateGraspPose stage which takes in a single target pose.
Using that template, I was able to easily create a custom stage which instead takes in a Pose array and generates solutions for all those poses as a starting point for planning.
I feel this is a common requirement for a lot of manipulation pipelines, and was wondering if we could have that as an added stage in the moveit core itself!
(FYI: I use the humble branch)
The text was updated successfully, but these errors were encountered: