Control Through Contact using Mixture of Deep Neural-Net Experts

Öz

We provide a data-driven control design framework for hybrid systems, with a special emphasis on contact-rich robotic systems. These systems exhibit continuous state flows and discrete state transitions, which are governed by distinct equations of motion. Hence, it may be impossible to design a single policy that can control the system in all modes. Typically, hybrid systems are controlled by multi-modal policies, each manually triggered based on observed states. However, as the number of potential contacts increase, the number of policies can grow exponentially and the control-switching scheme becomes too complicated to parameterize. To address this issue, we design contact-aware data-driven controllers given by deepnet mixture of experts (MoE). This architecture automatically learns switching-control scheme that can achieve the desired overall performance of the system, and a gating network, which determines the region of validity of each expert, based on the observed states.

Anahtar Kelimeler

• robotics
• nonlinear control
• machine learning

Kaynakça

1. [1] M. Cutkosky, “On grasp choice, grasp models, and the design of hands for manufacturing tasks,” IEEE Transactions on Robotics and Automation, vol. 5, no. 3, pp. 269–279, 1989.
2. [2] X. Cheng, E. Huang, Y. Hou, and M. T. Mason, “Contact mode guided motion planning for quasidynamic dexterous manipulation in 3d,” in 2022 International Conference on Robotics and Automation (ICRA), pp. 2730–2736, IEEE, 2022.
3. [3] F. Ruggiero, V. Lippiello, and B. Siciliano, “Nonprehensile dynamic manipulation: A survey,” IEEE Robotics and Automation Letters, vol. 3, no. 3, pp. 1711–1718, 2018.
4. [4] F. Ruggiero, A. Petit, D. Serra, A. C. Satici, J. Cacace, A. Donaire, F. Ficuciello, L. R. Buonocore, G. A. Fontanelli, V. Lippiello, et al., “Nonprehensile manipulation of deformable objects: Achievements and perspectives from the robotic dynamic manipulation project,” IEEE Robotics & Automation Magazine, vol. 25, no. 3, pp. 83–92, 2018.
5. [5] K. M. Lynch and T. D. Murphey, “Control of nonprehensile manipulation,” in Control problems in robotics, pp. 39–57, Springer, 2003.
6. [6] K. M. Lynch and M. T. Mason, “Dynamic nonprehensile manipulation: Controllability, planning, and experiments,” The International Journal of Robotics Research, vol. 18, no. 1, pp. 64–92, 1999.
7. [7] M. Erdmann, “An exploration of nonprehensile two-palm manipulation,” The International Journal of Robotics Research, vol. 17, no. 5, pp. 485– 503, 1998.
8. [8] M. Yashima, Y. Shiina, and H. Yamaguchi, “Randomized manipulation planning for a multi-fingered hand by switching contact modes,” in 2003 IEEE International Conference on Robotics and Automation (Cat. No. 03CH37422), vol. 2, pp. 2689–2694, IEEE, 2003.

9. [9] J. Z. Woodruff and K. M. Lynch, “Planning and control for dynamic, nonprehensile, and hybrid manipulation tasks,” in 2017 IEEE International Conference on Robotics and Automation (ICRA), pp. 4066–4073, IEEE, 2017.
10. [10] K. Lowrey, S. Kolev, J. Dao, A. Rajeswaran, and E. Todorov, “Reinforcement learning for non-prehensile manipulation: Transfer from simulation to physical system,” in 2018 IEEE International Conference on Simulation, Modeling, and Programming for Autonomous Robots (SIMPAR), pp. 35–42, IEEE, 2018.
11. [11] X. Zhang, M. Chang, P. Kumar, and S. Gupta, “Diffusion meets dagger: Supercharging eye-in-hand imitation learning,” arXiv preprint arXiv:2402.17768, 2024.
12. [12] C. M. Bishop, “Pattern recognition,” Machine learning, vol. 128, no. 9, 2006.
13. [13] T. H¨ark¨onen, S. Wade, K. Law, and L. Roininen, “Mixtures of gaussian process experts with smc2,” arXiv preprint arXiv:2208.12830, 2022.
14. [14] M. I. Jordan and R. A. Jacobs, “Hierarchical mixtures of experts and the em algorithm,” Neural computation, vol. 6, no. 2, pp. 181–214, 1994.
15. [15] Z. Chen, Y. Deng, Y. Wu, Q. Gu, and Y. Li, “Towards understanding mixture of experts in deep learning,” arXiv preprint arXiv:2208.02813, 2022.
16. [16] M. M. Zhang and S. A. Williamson, “Embarrassingly parallel inference for gaussian processes,” Journal of Machine Learning Research, 2019.
17. [17] L. V. Jospin, W. Buntine, F. Boussaid, H. Laga, and M. Bennamoun, “Hands-on bayesian neural networks–a tutorial for deep learning users,” arXiv preprint arXiv:2007.06823, 2020.
18. [18] S. Sharma, S. Sharma, and A. Athaiya, “Activation functions in neural networks,” Towards Data Sci, vol. 6, no. 12, pp. 310–316, 2017.
19. [19] D. Liberzon, Switching in systems and control, vol. 190. Springer, 2003.
20. [20] R. Goebel, R. G. Sanfelice, and A. R. Teel, “Hybrid dynamical systems,” IEEE control systems magazine, vol. 29, no. 2, pp. 28–93, 2009.
21. [21] B. Brogliato and B. Brogliato, Nonsmooth mechanics. Springer, 1999.
22. [22] S. Ross, G. J. Gordon, and J. A. Bagnell, “No-regret reductions for imitation learning and structured prediction,” in In AISTATS, Citeseer, 2011.
23. [23] J. Revels, M. Lubin, and T. Papamarkou, “Forward-mode automatic differentiation in julia,” arXiv preprint arXiv:1607.07892, 2016.
24. [24] D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014.
25. [25] D.-A. Clevert, T. Unterthiner, and S. Hochreiter, “Fast and accurate deep network learning by exponential linear units (elus),” arXiv preprint arXiv:1511.07289, 2015.
26. [26] J. J. Moreau, “Unilateral contact and dry friction in finite freedom dynamics,” Nonsmooth mechanics and Applications, pp. 1–82, 1988.
27. [27] C. Glocker and C. Studer, “Formulation and preparation for numerical evaluation of linear complementarity systems in dynamics,” Multibody System Dynamics, vol. 13, pp. 447–463, 2005.
28. [28] V. Acary and B. Brogliato, Numerical methods for nonsmooth dynamical systems: applications in mechanics and electronics. Springer Science & Business Media, 2008.
29. [29] Quanser, Linear Servo Base Unit with Inverted Pendulum. Apr 2021.
30. [30] M. Spong, S. Hutchinson, and M. Vidyasagar, Robot Modeling and Control. Wiley, 2020.
31. [31] H. Khalil, Nonlinear Control. Always Learning, Pearson, 2015.