First of all, thank you for releasing this package. We are currently testing different kinematics and how they behave with different paths and it seems like you have done a really good job!
I do have some questions for you:
Software limits. It seems the motion planner detects there will be a software limit overrun quite late even if the position operations were buffered a long time ago. The planner seems independent of the "software error reaction deceleration/distance", because even if I change these the error is always generated at the exact same position. I was kinda hoping that the "lookahead" would be as long as the axis requiring the longest stop distance at the currently executed speed. Any comments on this?
Blending. This is more a theoretical question and to understand how the underlying mechanics in the robotic engine work. Let's say I have hundreds of short linear movements (minimum 3mm) describing a long freeshape curve several meters long. Will the new robotic routines be able to plan and execute these with a vector speed of 200mm/s given that individual dynamic limits are not exceeded? I'm asking because this means covering about 70 linear operations per second, and I don't know if this can be buffered properly etc.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
We are aware that software limits are not always detected as early as possible and are currently working to improve this. There are two things we will improve: software limits will be detected before they are actually exceeded. As soon as we notice that they would be exceeded if continuing the current movement, we create a limit error and brake in time to not violate the limit. (SM-65, already done.) Second, the robotics planner will be improved to perform a path accurate stop before a software limit is violated. (SM-767, currently in progress)
Blending for linear movements in Robotics is similar to the way it is done in our CNC with SMC_SmoothPath. If you command the linear movements in advance, blending elements will be inserted between the linear elements. The second linear element of two blended elements will be replaced by a blending element up to half its length. This limits the amount of blending that can be done, since each blending element is rather short. However, the planning of the velocity profile on the path is done on the blended path and should have no problem to achieve a high path velocity.
(Note that we implemented the CNC function block SMC_SmoothMerge for the scenario of many very short linear elements, as often produced by CAD/CAM systems, e.g. with length ~ 0.1 mm. SMC_SmoothMerge will try to fit as many consecutive linear elements by a higher order polynomial (in the least squares sense), until a given error tolerance is exceeded. We plan to combine CNC and robotics in the future, most likely by means of MC_MovePath. This will make it possible to use SMC_SmoothMerge and other CNC function blocks and to execute the path by a Robotics axis group.)
BR
Edwin
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
First of all, thank you for releasing this package. We are currently testing different kinematics and how they behave with different paths and it seems like you have done a really good job!
I do have some questions for you:
Software limits. It seems the motion planner detects there will be a software limit overrun quite late even if the position operations were buffered a long time ago. The planner seems independent of the "software error reaction deceleration/distance", because even if I change these the error is always generated at the exact same position. I was kinda hoping that the "lookahead" would be as long as the axis requiring the longest stop distance at the currently executed speed. Any comments on this?
Blending. This is more a theoretical question and to understand how the underlying mechanics in the robotic engine work. Let's say I have hundreds of short linear movements (minimum 3mm) describing a long freeshape curve several meters long. Will the new robotic routines be able to plan and execute these with a vector speed of 200mm/s given that individual dynamic limits are not exceeded? I'm asking because this means covering about 70 linear operations per second, and I don't know if this can be buffered properly etc.
Hi Kim,
We are aware that software limits are not always detected as early as possible and are currently working to improve this. There are two things we will improve: software limits will be detected before they are actually exceeded. As soon as we notice that they would be exceeded if continuing the current movement, we create a limit error and brake in time to not violate the limit. (SM-65, already done.) Second, the robotics planner will be improved to perform a path accurate stop before a software limit is violated. (SM-767, currently in progress)
Blending for linear movements in Robotics is similar to the way it is done in our CNC with SMC_SmoothPath. If you command the linear movements in advance, blending elements will be inserted between the linear elements. The second linear element of two blended elements will be replaced by a blending element up to half its length. This limits the amount of blending that can be done, since each blending element is rather short. However, the planning of the velocity profile on the path is done on the blended path and should have no problem to achieve a high path velocity.
(Note that we implemented the CNC function block SMC_SmoothMerge for the scenario of many very short linear elements, as often produced by CAD/CAM systems, e.g. with length ~ 0.1 mm. SMC_SmoothMerge will try to fit as many consecutive linear elements by a higher order polynomial (in the least squares sense), until a given error tolerance is exceeded. We plan to combine CNC and robotics in the future, most likely by means of MC_MovePath. This will make it possible to use SMC_SmoothMerge and other CNC function blocks and to execute the path by a Robotics axis group.)
BR
Edwin
Thank you Edwin, very good answers, no further questions!