Best paper award provided by the TC on Model-based Optimization for Robotics

[jcarpent]

Dear Pinocchio community,

We are thrilled to announce that one of our recent algorithms for accounting constraints in forward dynamics with the lowest complexity (aka Constrained Articulated Body Algorithm) has received the best paper award from TC on Model-based Optimization for Robotics.

This is a nice and encouraging recognition of our recent efforts to push further the limits regarding efficiency, versatility, and complexity of rigid-body dynamics algorithms. You can freely download the related paper here and its extension for closed-loop constraints here.

Kudos @AjSat for this nice work!
And more to come very shortly, hopefully!

[DjoleMNE]

Hi @jcarpent and @AjSat ,

Thank you so much for sharing this with us!

These algorithms are so useful! Robotics community is catching up with them these days. And you guys did an amazing job in extending them further and developing new solutions in this field!

Here is a further list of resent related publications on these type of algorithms:

• Sathya, A. S., Bruyninckx, H., Decre, W., & Pipeleers, G. (2024). Efficient Constrained Dynamics Algorithms Based on an Equivalent LQR Formulation Using Gauss' Principle of Least Constraint. IEEE Transactions on Robotics, 40, 729-749.

• C. Mastalli, S. P. Chhatoi, T. Corbéres, S. Tonneau and S. Vijayakumar, "Inverse-Dynamics MPC via Nullspace Resolution," in IEEE Transactions on Robotics, vol. 39, no. 4, pp. 3222-3241, Aug. 2023 (Authors also use Gauss' principle of least constraints in this publication)

• Justin Carpentier, Quentin Le Lidec, Louis Montaut. From Compliant to Rigid Contact Simulation: a Unified and Efficient Approach. 20th edition of the “Robotics: Science and Systems” (RSS) Conference, Jul 2024, Delft, Netherlands.

• Carpentier, Justin, Rohan Budhiraja, and Nicolas Mansard. "Proximal and sparse resolution of constrained dynamic equations." In Robotics: Science and Systems 2021.

• S. Schneider and H. Bruyninckx, "Exploiting linearity in dynamics solvers for the design of composable robotic manipulation architectures," 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). (Authors also use Gauss' principle of least constraints in this publication)

• Yang L, Xue S, Yao W. Application of Gauss principle of least constraint in multibody systems with redundant constraints. Proceedings of the Institution of Mechanical Engineers, Part K: Journal of Multi-body Dynamics. 2020;235(1):150-163.

There are of course older publications from Todorov, Erez and Tassa, Kheddar, Udwadia and Kalaba, Khatib and Bruyninckx, Popov and Vereshchagin, etc.

I used the original version of the Popov-Vereshchagin algorithm with static friction in 2018, and it allowed me to create a very nice impedance/force controller with it for manipulators:

• Static-friction extension
• Lazy manipulator control

Back then, I used KDL's implementation of basic PV solver, but I would like to start using Pinocchio's advanced versions of this algorithm.

How's the progress of implementation of these algorithms in the Pinocchio and Crocoddyl library?

I found out there are PV solver files in Pinocchio, but because documentation about this solver's implementation is limited, can you please give us more infomation about the implementation in Pinocchio and Crocoddyl? Is the extended version for closed-loop systems also implemented, or just for serial/tree chains? Does the solver include dissipative forces in the computations? How about analytical derivatives of PV solver, are they also implemented? Examples on how to use API of these solvers? And so on...

If I missed some documentation on this, I apologize in advance!

Thanks a lot in advance! This information would be so important for us who would like to start using these algorithms more extensively in the future. Since the algorithms are very advanced, it would be hard for robotics community to use them in Pinocchio and Crocoddyl, if a good documentation and practical examples are not there. Feel free to write us as much as you can.

If necessary, we can create a new thread here on GitHub-discussions for discussing the implementation and documentation of these algorithms in more detail.

Best regards,
Djordje

P.S.

@AjSat , we met a few years ago in Leuven, on KU Leuven Arenberg campus during a BBQ gathering. That's when I got to know that you implemented PV solver in Pinocchio :D
Hopefully, we will meet again sometime soon!

[cmastalli]

Hey @DjoleMNE!

Regarding Crocoddyl development, we've an ongoing effort that integrates closed-loop constraints. However, we will release this after we have finished other new features in Crocoddyl. Please stay tuned!

Best,
Carlos

[DjoleMNE]

@cmastalli ,

that's awesome news, thanks a lot for the feedback! 👌😁🙏

[jcarpent]

@DjoleMNE, thank you for your interest in our recent work on accurately and efficiently handling closed-loop mechanisms for robot control and simulation.
It is relatively rare to have such precious feedback. Thank you.

To move forward on this discussion, I suggest organizing a dedicated call between you and us (Ajay and me). We will be able to show you our latest, yet unpublished work, on the topic, and discuss integration within Crocoddyl or Aligator, our new constrained trajectory optimization solver, fully parallelizable and with nice features for online nonlinear model predictive control (see here and here for paper references).

[DjoleMNE]

@jcarpent , thanks a lot for the call invite!

That would be very helpful! Looking forward to this discussion!

@AjSat and I are connected on LinkedIn. You guys can send me there your availability for the call, and we can arrange everything from there. Let me know if that is fine with you.

Thanks a lot again 😄

[jcarpent]

I don’t have LinkedIn. @DjoleMNE Could you send me an email instead?

[DjoleMNE]

Yes, of course @jcarpent . I just sent you an email.