bmferguson
Member
Hi all,
I would appreciate some advice on which path to take in order to achieve my goal. I want to load a robot that I designed in Alibre Design and train it using Reinforcement Learning. For that, I ultimately need to express it in URDF or SDF format.
I need to make an informed decision before I spend too much time and/or money on something that may not work, and I hope that you might have some helpful insights.
Here are my options as I see them:
1) Write an Alibre tool using the API that allows me to convert directly into URDF or SDF format.
- Pro(s): It would remove many steps for subsequent creations
- Con(s): It would be yet another language/framework I would have to master, and the information I have so far suggests that it might not be that straightforward to capture constraint information.
2) Wait for Alibre to write something.
- Pro(s): I wouldn't have to write it, and I could concentrate on the many other balls I have to juggle. Also, it would undoubtedly be a better product.
- Con(s): I could be waiting for something that may never materialize, particularly because I have no idea whether or not Alibre has an interest in tapping into a market (robotics) valued at US$ 26.52 Billion globally in 2022 with a CAGR of 10.5% between 2023 and 2030....hint, hint .
3) Pay for the standalone KeyShot Pro and attempt a conversion from Alibre Expert -> KeyShot Pro -> USD -> SDF -> PyBullet
- Pro(s): It's not the end of the world cost-wise (assuming it's a viable option), and maybe the textures would be preserved in the final model.
- Con(s): It's not clear to me whether or not any of the individual parts or link/constraint information would survive the conversion, and significant manual work could still be required.
4) Switch to or also use SolidWorks, OnShape, or Fusion 360 (I really don't want to do this)
- Pro(s): The functionality currently exists. The thought of creating something and pressing a button or defining things interactively in order to move to the next step seems like it would help the flow quite nicely.
- Con(s): Cost, subscription model, online storage, vendor lock, etc...
5) Painstakingly construct the URDF or SDF manually.
- Pro(s): No more $
- Con(s): Significant time and effort. My design isn't the most simple thing, and the effort to put all of the pieces and associations together is daunting.
6) Blender + Phobos?
Does anyone see any other options (such as some other conversion path), or have some advice/suggestions for me?
Thanks in advance,
Bruce
I would appreciate some advice on which path to take in order to achieve my goal. I want to load a robot that I designed in Alibre Design and train it using Reinforcement Learning. For that, I ultimately need to express it in URDF or SDF format.
I need to make an informed decision before I spend too much time and/or money on something that may not work, and I hope that you might have some helpful insights.
Here are my options as I see them:
1) Write an Alibre tool using the API that allows me to convert directly into URDF or SDF format.
- Pro(s): It would remove many steps for subsequent creations
- Con(s): It would be yet another language/framework I would have to master, and the information I have so far suggests that it might not be that straightforward to capture constraint information.
2) Wait for Alibre to write something.
- Pro(s): I wouldn't have to write it, and I could concentrate on the many other balls I have to juggle. Also, it would undoubtedly be a better product.
- Con(s): I could be waiting for something that may never materialize, particularly because I have no idea whether or not Alibre has an interest in tapping into a market (robotics) valued at US$ 26.52 Billion globally in 2022 with a CAGR of 10.5% between 2023 and 2030....hint, hint .
3) Pay for the standalone KeyShot Pro and attempt a conversion from Alibre Expert -> KeyShot Pro -> USD -> SDF -> PyBullet
- Pro(s): It's not the end of the world cost-wise (assuming it's a viable option), and maybe the textures would be preserved in the final model.
- Con(s): It's not clear to me whether or not any of the individual parts or link/constraint information would survive the conversion, and significant manual work could still be required.
4) Switch to or also use SolidWorks, OnShape, or Fusion 360 (I really don't want to do this)
- Pro(s): The functionality currently exists. The thought of creating something and pressing a button or defining things interactively in order to move to the next step seems like it would help the flow quite nicely.
- Con(s): Cost, subscription model, online storage, vendor lock, etc...
5) Painstakingly construct the URDF or SDF manually.
- Pro(s): No more $
- Con(s): Significant time and effort. My design isn't the most simple thing, and the effort to put all of the pieces and associations together is daunting.
6) Blender + Phobos?
Does anyone see any other options (such as some other conversion path), or have some advice/suggestions for me?
Thanks in advance,
Bruce
Last edited: