What is a good local description of contact dynamics for contact-rich manipulation, and where can we trust this local description? While many approaches often rely on the Taylor approximation of dynamics with an ellipsoidal trust region, we argue that such approaches are fundamentally inconsistent with the unilateral nature of contact. As a remedy, we present the contact trust region (CTR), which captures the unilateral nature of contact while remaining efficient for computation. With CTR, we first develop a model-predictive control (MPC) algorithm capable of synthesizing local contact-rich plans. Then, we extend this capability to plan globally by stitching together local MPC plans, enabling efficient and dexterous contact-rich manipulation. To verify the performance of our method, we perform comprehensive evaluations, both in high-fidelity simulation and on hardware, on two contact-rich systems: a planar IiwaBimanual system and a 3D AllegroHand system. On both systems, our method offers a significantly lower-compute alternative to existing RL-based approaches to contact-rich manipulation. In particular, our Allegro in-hand manipulation policy, in the form of a roadmap, takes fewer than 10 minutes to build offline on a standard laptop using just its CPU, with online inference taking just a few seconds. Experiment data, video and code are available at ctr.theaiinstitute.com.
AydinogluAPosaM (2022) Real-time multi-contact model predictive control via admm. In: 2022 International Conference on Robotics and Automation (ICRA), 3414–3421. DOI: 10.1109/ICRA46639.2022.9811957.
3.
AydinogluASiegPPreciadoVM, et al. (2022) Stabilization of complementarity systems via contact-aware controllers. IEEE Transactions on Robotics38(3): 1735–1754.
4.
BarreirosJAÖzgün ÖnolAZhangM, et al. (2025) Learning contact-rich whole-body manipulation with example-guided reinforcement learning. Science Robotics10(105): eads6790.
5.
BlackKBrownNDriessD, et al. (2024) π0: a vision-language-action flow model for general robot control. arXiv preprint arXiv:2410.24164.
6.
BoydSVandenbergheL (2004) Convex Optimization. Cambridge University Press.
7.
CastroAHanXMasterjohnJ (2023) A theory of irrotational contact fields. arXiv preprint arXiv:2312.03908.
8.
Chavan-DafleNHolladayRRodriguezA (2020) Planar in-hand manipulation via motion cones. The International Journal of Robotics Research39(2-3): 163–182.
9.
ChenTXuJAgrawalP (2022) A system for general in-hand object re-orientation. In: FaustAHsuDNeumannG (eds). Proceedings of the 5th Conference on Robot Learning, Proceedings of Machine Learning Research. PMLR, Vol. 164, 297–307. https://proceedings.mlr.press/v164/chen22a.html
10.
ChengXHuangEHouY, et al. (2022) Contact mode guided motion planning for quasidynamic dexterous manipulation in 3d. In: 2022 International Conference on Robotics and Automation (ICRA), 2730–2736. DOI: 10.1109/ICRA46639.2022.9811872.
11.
ChengXPatilSTemelZ, et al. (2024) Enhancing dexterity in robotic manipulation via hierarchical contact exploration. IEEE Robotics and Automation Letters9(1): 390–397.
12.
ChiCXuZFengS, et al. (2024) Diffusion policy: visuomotor policy learning via action diffusion. The International Journal of Robotics Research.
13.
ChosetHLynchKMHutchinsonS, et al. (2005) Principles of Robot Motion: Theory, Algorithms, and Implementations. MIT press.
DaiH (2016) Robust multi-contact Dynamical Motion Planning Using Contact Wrench Set. Massachusetts Institute of Technology. PhD Thesis.
16.
DaiHMajumdarATedrakeR (2018) Synthesis and Optimization of Force Closure Grasps via Sequential Semidefinite Programming. Springer International Publishing, 285–305.
17.
DingYPandalaAParkHW (2019) Real-time model predictive control for versatile dynamic motions in quadrupedal robots. In: 2019 International Conference on Robotics and Automation (ICRA), 8484–8490. DOI: 10.1109/ICRA.2019.8793669.
18.
ErdmannM (1994) On a representation of friction in configuration space. The International Journal of Robotics Research13(3): 240–271.
19.
FerrariCCannyJ (1992) Planning optimal grasps. Proceedings 1992 IEEE International Conference on Robotics and Automation3: 2290–2295.
20.
FiaccoAV (1983) Introduction to sensitivity and stability analysis in non linear programming.
21.
FreemanCDFreyERaichukA, et al. (2021) Brax - a differentiable physics engine for large scale rigid body simulation. https://github.com/google/brax
22.
GiorgiGZuccottiC (2018) A Tutorial on Sensitivity and Stability in Nonlinear Programming and Variational Inequalities Under Differentiability Assumptions. University of Pavia, Department of Economics and Management, DEM Working Papers Series 159.
23.
GraesdalBPChiaSYCMarcucciT, et al. (2024) Towards tight convex relaxations for contact-rich manipulation. arXiv preprint arXiv:2402.10312.
24.
HanLTrinkleJLiZ (2000) Grasp analysis as linear matrix inequality problems. IEEE Transactions on Robotics and Automation16(6): 663–674.
25.
HanXMasterjohnJCastroA (2023) A convex formulation of frictional contact between rigid and deformable bodies. IEEE Robotics and Automation Letters8(10): 6219–6226.
26.
HandaAAllshireAMakoviychukV, et al. (2022) Dextreme: transfer of agile in-hand manipulation from simulation to reality. arXiv.
27.
HowellTGileadiNTunyasuvunakoolS, et al. (2022) Predictive sampling: real-time behaviour synthesis with Mujoco. https://arxiv.org/abs/2212.00541
HowellTACleac’hSLSinghS, et al. (2023b) Trajectory optimization with optimization-based dynamics. https://arxiv.org/abs/2109.04928
30.
HuangEChengXMasonMT (2020) Efficient contact mode enumeration in 3d.
31.
JinW (2024) Complementarity-free multi-contact modeling and optimization for dexterous manipulation. https://arxiv.org/abs/2408.07855
32.
KavrakiLSvestkaPLatombeJC, et al. (1996) Probabilistic roadmaps for path planning in high-dimensional configuration spaces. IEEE Transactions on Robotics and Automation12(4): 566–580.
33.
KhandateGSaidiTLShangS, et al. (2024) R times r: rapid exploration for reinforcement learning via sampling-based reset distributions and imitation pre-training. Autonomous Robots48(7): 17.
34.
KongNJLiCCouncilG, et al. (2023) Hybrid ilqr model predictive control for contact implicit stabilization on legged robots. IEEE Transactions on Robotics39(6): 4712–4727.
KurtzVCastroAÖnolAÖ, et al. (2023) Inverse dynamics trajectory optimization for contact-implicit model predictive control. The International Journal of Robotics Research.
37.
LBM Team TRI (2025) A careful examination of large behavior models for multitask dexterous manipulation URL. https://arxiv.org/abs/2507.05331
38.
Le Cleac’hSSchwagerMManchesterZ, et al. (2023) Single-level differentiable contact simulation. IEEE Robotics and Automation Letters8(7): 4012–4019.
39.
Le Cleac’hSHowellTAYangS, et al. (2024) Fast contact-implicit model predictive control. IEEE Transactions on Robotics40: 1617–1629.
LiWTodorovE (2004) Iterative linear quadratic regulator design for nonlinear biological movement systems. In: First International Conference on Informatics in Control, Automation and Robotics. SciTePress, Vol. 2, 222–229.
42.
LiAHCulbertsonPBurdickJW, et al. (2023) Frogger: fast robust grasp generation via the min-weight metric. In: 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 6809–6816. DOI: 10.1109/IROS55552.2023.10341806.
43.
LiAHCulbertsonPKurtzV, et al. (2024) Drop: dexterous reorientation via online planning. arXiv preprint arXiv:2409.14562 Available at: https://arxiv.org/abs/2409.14562
44.
LiYHanHKangS, et al. (2025) On the surprising robustness of sequential convex optimization for contact-implicit motion planning. arXiv preprint arXiv:2502.01055.
45.
LynchKParkF (2017) Modern Robotics: Mechanics, Planning, and Control. Cambridge University Press.
MagnusJRNeudeckerH (2019) Matrix Differential Calculus with Applications in Statistics and Econometrics. John Wiley & Sons.
48.
MarcucciTTedrakeR (2019) Mixed-integer formulations for optimal control of piecewise-affine systems. In: Proceedings of the 22nd ACM International Conference on Hybrid Systems: Computation and Control, 230–239.
49.
MarcucciTDeitsRGabicciniM, et al. (2017) Approximate hybrid model predictive control for multi-contact push recovery in complex environments. In: 2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids), 31–38. DOI: 10.1109/HUMANOIDS.2017.8239534.
50.
MarcucciTUmenbergerJParriloP, et al. (2024) Shortest paths in graphs of convex sets. SIAM Journal on Optimization34(1): 507–532.
51.
MasonMT (1986) Mechanics and planning of manipulator pushing operations. The International Journal of Robotics Research5(3): 53–71.
52.
MasonMTSalisburyJK (1985) Robot Hands and the Mechanics of Manipulation. MIT Press.
53.
MastalliCBudhirajaRMerktW, et al. (2020) Crocoddyl: an efficient and versatile framework for multi-contact optimal control. In: 2020 IEEE International Conference on Robotics and Automation (ICRA), 2536–2542. DOI: 10.1109/ICRA40945.2020.9196673.
54.
MoreJJ (1993) Generalizations of the trust region problem. Optimization Methods and Software2(3-4): 189–209.
55.
MoréJJSorensenDC (1983) Computing a trust region step. SIAM Journal on Scientific and Statistical Computing4(3): 553–572.
OnolAOLongPPadırT (2019) Contact-implicit trajectory optimization based on a variable smooth contact model and successive convexification. In: 2019 International Conference on Robotics and Automation (ICRA), 2447–2453. DOI: 10.1109/ICRA.2019.8794250.
OrricoCAYangBKrishnamoorthyD (2024) On building myopic mpc policies using supervised learning. arXiv preprint arXiv:2401.12546.
63.
PangTTedrakeR (2021) A convex quasistatic time-stepping scheme for rigid multibody systems with contact and friction. In: 2021 IEEE International Conference on Robotics and Automation (ICRA), 6614–6620. DOI: 10.1109/ICRA48506.2021.9560941.
64.
PangTTedrakeR (2022) Easing reliance on collision-free planning with contact-aware control. In: 2022 International Conference on Robotics and Automation (ICRA). IEEE, 8375–8381.
65.
PangTSuhHJTYangL, et al. (2023) Global planning for contact-rich manipulation via local smoothing of quasi-dynamic contact models. IEEE Transactions on Robotics39(6): 4691–4711.
66.
PengXBAbbeelPLevineS, et al. (2018) Deepmimic: example-guided deep reinforcement learning of physics-based character skills. ACM Transactions on Graphics37(4): 1–14.
67.
PengXBMaZAbbeelP, et al. (2021) Amp: adversarial motion priors for stylized physics-based character control. ACM Transactions on Graphics40(4): 1–20.
68.
PetersenMTedrakeR (2023) Growing convex collision-free regions in configuration space using nonlinear programming. arXiv preprint arXiv:2303.14737.
69.
PosaMCantuCTedrakeR (2014) A direct method for trajectory optimization of rigid bodies through contact. The International Journal of Robotics Research33(1): 69–81.
70.
PrattichizzoDTrinkleJC (2008) Grasping. Springer Berlin Heidelberg, 671–700. DOI: 10.1007/978-3-540-30301-5_29.
71.
RajeswaranAKumarVGuptaA, et al. (2018) Learning complex dexterous manipulation with deep reinforcement learning and demonstrations. https://arxiv.org/abs/1709.10087
72.
SchwartzA (2011) Mathematical Programs with Complementarity Constraints: Theory, Methods, and Applications. University of Wurzburg. PhD Thesis.
73.
ShiraiYZhaoTSuhHT, et al. (2025) Is linear feedback on smoothed dynamics sufficient for stabilizing contact-rich plans? In: 2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 11926–11932.
74.
SongYRomeroAMüllerM, et al. (2023) Reaching the limit in autonomous racing: optimal control versus reinforcement learning. Science Robotics8(82): eadg1462.
75.
SorensenDC (1982) Newton’s method with a model trust region modification. SIAM Journal on Numerical Analysis19(2): 409–426.
76.
StewartDETrinkleJC (2000) An implicit time-stepping scheme for rigid body dynamics with coulomb friction. Proceedings 2000 ICRA. Millennium Conference. IEEE International Conference on Robotics and Automation. Symposia Proceedings (Cat. No.00CH37065)1: 162–169. https://api.semanticscholar.org/CorpusID:10920958
77.
SuhHJSimchowitzMZhangK, et al. (2022a) Do differentiable simulators give better policy gradients? In: ChaudhuriKJegelkaSSongL, et al. (eds). Proceedings of the 39th International Conference on Machine Learning, Proceedings of Machine Learning Research. PMLR, Vol. 162, 20668–20696. https://proceedings.mlr.press/v162/suh22b.html
78.
SuhHJTPangTTedrakeR (2022b) Bundled gradients through contact via randomized smoothing. IEEE Robotics and Automation Letters7(2): 4000–4007.
79.
TassaYErezTTodorovE (2012) Synthesis and stabilization of complex behaviors through online trajectory optimization. In: 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, 4906–4913. DOI: 10.1109/IROS.2012.6386025.
TedrakeRThe Drake Development Team (2019) Drake: Model-based design and verification for robotics. https://drake.mit.edu
82.
TodorovE (2014) Convex and analytically-invertible dynamics with contacts and constraints: theory and implementation in mujoco. In: 2014 IEEE International Conference on Robotics and Automation (ICRA), 6054–6061. DOI: 10.1109/ICRA.2014.6907751.
83.
TodorovEErezTTassaY (2012) Mujoco: a physics engine for model-based control. In: 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, 5026–5033. DOI: 10.1109/IROS.2012.6386109.
84.
WensingPMPosaMHuY, et al. (2024) Optimization-based control for dynamic legged robots. IEEE Transactions on Robotics40: 43–63.
85.
WerlingKOmensDLeeJ, et al. (2021) Fast and feature-complete differentiable physics for articulated rigid bodies with contact. arXiv preprint arXiv:2103.16021.
86.
YangLSuhHJTZhaoT, et al. (2025) Physics-driven data generation for contact-rich manipulation via trajectory optimization. https://arxiv.org/abs/2502.20382
87.
ZhangSJinWWangZ (2023) Adaptive barrier smoothing for first-order policy gradient with contact dynamics. In: KrauseABrunskillEChoK, et al. (eds). Proceedings of the 40th International Conference on Machine Learning, Proceedings of Machine Learning Research. PMLR, Vol. 202, 41219–41243. https://proceedings.mlr.press/v202/zhang23s.html
88.
ZhangTTPfrommerDMatniN, et al. (2025) Imitation learning in continuous action spaces: mitigating compounding error without interaction. arXiv preprint arXiv:2507.09061.
89.
ZhongYDHanJBrikisGO (2022) Differentiable physics simulations with contacts: do they have correct gradients w.r.t. position, velocity and control?https://arxiv.org/abs/2207.05060
90.
ZhuHZhaoTNiX, et al. (2025) Should we learn contact-rich manipulation policies from sampling-based planners?IEEE Robotics and Automation Letters10(6): 6248–6255.
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
