Robotics StackExchange | Archived questions

Moveit linear motion pick & place

Using the moveit pick&place API from python I am getting too many suboptimal paths. I tried several planners such as EST, RRT and some others.

I would like to use the pick&place api but working in a much more cartesian-linear behavior. How could I do that?

I could manually generate the trajectory interpolating from two joint configurations or from( two cartesian poses + iterative IK interpolation). However these approach would have several drawbacks that I would like to avoid: - need to implement the pick and place operations (approach, retreive, etc.) - no collision checking - need to implement joint limits checking

Asked by Pablo Iñigo Blasco on 2018-09-16 05:44:47 UTC

Comments

Answers

One option is to use the computeCartesianPath() method which is a straight forward interpolation in Cartesian space followed by the corresponding IK operations. The other option is the descartes planner which uses a graph-search based approach in order to find an optimal path.

Asked by jrgnicho on 2018-09-17 12:58:28 UTC

Comments

As far as I know computeCartesianPath is a funcionality outside the pick and place api. Using that I would need to implement all the steps of pick an place by my own. Am I right?

The question is: how can I keep the pick & place with obstacle avoidance functionality with linear motions?

Asked by Pablo Iñigo Blasco on 2018-09-18 18:02:18 UTC

Descartes might be closer, but it does not look totaly integrated in moveit and it is not in the official repositories :S

Asked by Pablo Iñigo Blasco on 2018-09-18 18:03:15 UTC