Learning Stable Nonlinear Dynamical Systems With Gaussian Mixture Models
Open accessThe DOI matches its Crossref record; 41 of the 43 references it cites were found in a registry or an index.
Abstract
Abstract—This paper presents a method for learning discrete robot motions from a set of demonstrations. We model a motion as a nonlinear autonomous (i.e. time-invariant) Dynamical System (DS), and define sufficient conditions to ensure global asymptotic stability at the target. We propose a learning method, called Stable Estimator of Dynamical Systems (SEDS), to learn the parameters of the DS to ensure that all motions follow closely the demonstrations while ultimately reaching in and stopping at the target. Time-invariance and global asymptotic stability at the target ensures that the system can respond immediately and appropriately to perturbations encountered during the motion. The method is evaluated through a set of robot experiments and on a library of human handwriting motions.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“In this model the convergence of all trajectories to the target is guaranteed, but the final velocity at the target is always zero (i.e. the target is an asymptotically stable point).”
From Learning to Play Minigolf: A Dynamical System-Based Approach (2012) · cites this paper as [2]
In the citing paper
“The resulting model from SEDS is globally asymptotically stable, i.e. starting from any point in the space, all trajectories converge to the origin.”
From Learning to Play Minigolf: A Dynamical System-Based Approach (2012) · cites this paper as [2]
In the citing paper
“The optimization constraints given by Eq. (10) ensure the global asymptotic stability of the function f̂(x;θ) (see [2]).”
From Learning to Play Minigolf: A Dynamical System-Based Approach (2012) · cites this paper as [2]
In the citing paper
“Gaussian distributions are used in PoMPs [114] and SEDs [115] for better generalisation behaviour.”
From Data-Driven Robotic Manipulation of Cloth-like Deformable Objects: The Present, Challenges and Future Prospects (2023) · cites this paper as [115]
In the citing paper
“We use the dynamical systems approach to encode motion from several demonstrations. This enables a rapid and reactive adaptation of the arm motion in the presence of sensor uncertainty.”
From Catching Objects in Flight (2014) · cites this paper as [20]
1 more claim about this paper
Create a free account to see every claim other papers make about this one.
Citations
Cites
A new look at the statistical model identification
Akaike, H. (1974) · IEEE Transactions on Automatic Control · doi:10.1109/tac.1974.1100705 · Cited by 51,879
Maximum Likelihood from Incomplete Data Via the EM Algorithm
Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977) · Journal of the Royal Statistical Society Series B: Statistical Methodology · doi:10.1111/j.2517-6161.1977.tb01600.x · Cited by 50,444
Estimating the Dimension of a Model
Schwarz, G. (1978) · The Annals of Statistics · doi:10.1214/aos/1176344136 · Cited by 47,033
Slotine, J.-J. E., & Li, W. (1991) · CERN Document Server (European Organization for Nuclear Research) · Cited by 18,975
Gaussian Processes for Machine Learning
Rasmussen, C. E., & Williams, C. K. I. (2005) · doi:10.7551/mitpress/3206.001.0001 · Cited by 11,705
McLachlan, G., & Peel, D. (2000) · Wiley Series in Probability and Statistics · doi:10.1002/0471721182 · Cited by 10,991
Nonlinear programming: theory and algorithms
(1993) · Computers & Operations Research · doi:10.1016/0305-0548(93)90042-h · Cited by 5,779
ANew Look at the Statistical Model Identification
AKAlKE, H. (2011) · Cited by 4,177
Nonlinear Programming: Theory and Algorithms
Bazaraa, M. S., Sherali, H. D., & Shetty, C. M. (2006) · Wiley eBooks · doi:10.1002/0471787779 · Cited by 2,369
Sparse Gaussian Processes using Pseudo-inputs
Snelson, E., & Ghahramani, Z. (2005) · Cited by 1,415
Research works printed but not linked: 1
- schaal, 2002, Scalable locally weighted statistical techniques for real time robot learning, Appl Intell Spec Issue Scalable Robot Appl Neural Netw, 17, 49
Cited by
Items 501–600 of 772
Chen, L., Qiu, J., Zou, X., & Cheng, H. (2023) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48891.2023.10160949 · Cited by 3
“of Dynamical Systems (SEDS) [13], Probabilistic Movement Primitives (ProMP) [14] and Kernelized Movement Primitives (KMP) [15], etc.”
BackgroundZhai, B., Zhang, E., Li, B., & Fang, X. (2023) · Journal of Mechanisms and Robotics · doi:10.1115/1.4062985 · Cited by 3
Augmented-SVM: Automatic space partitioning for combining multiple non-linear dynamics
Shukla, A., & Billard, A. (2012) · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · Cited by 7
Self-citationA Segmentation Scheme for CLF Dynamic Movement Control Applied to Robotic Handwriting
Göttsch, P., Olschewski, R., & Werner, H. (2017) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2017.08.1818 · Cited by 3
“Another approach was presented in Khansari-Zadeh and Billard (2011), where stability conditions were provided when using Gaussian Mixture Model (GMM) for modelling the DS.”
BackgroundMethodInfluentialObstacle Avoidance Methods Based on Geometric Information Under the DMPs Framework
Shi, M., Gao, Y., Ti, B., & Zhao, J. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-89098-8_6 · Cited by 3
A Bayesian tracker for synthesizing mobile robot behaviour from demonstration
Magnenat, S., & Colas, F. (2021) · Autonomous Robots · doi:10.1007/s10514-021-10019-4 · Cited by 3
“Another approach to model the motion is to consider dynamical systems (Khansari-Zadeh and Billard 2011; Pastor et al. 2013). For instance, Pastor et al. (2013) have described skills as a combination of dynamic movement primitives.”
BackgroundImpedance adjustment for learning human variable impedance skill
Wu, R., Zhang, H., Zhu, Y., Zang, X., & Zhao, J. (2023) · Mechatronics · doi:10.1016/j.mechatronics.2023.103036 · Cited by 6
Safe Execution of Learned Orientation Skills with Conic Control Barrier Functions
Shen, Z., Saveriano, M., Abu‐Dakka, F. J., & Haddadin, S. (2024) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra57147.2024.10611325 · Cited by 9
“Among LfD approaches, DSs are attractive due to their capability to generate real-time motions and converge toward a predefined target [2].”
BackgroundMethodSensorimotor Learning With Stability Guarantees via Autonomous Neural Dynamic Policies
Totsila, D., Chatzilygeroudis, K. I., Modugno, V., Hadjivelichkov, D., & Kanoulas, D. (2025) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3524878 · Cited by 6
“For example, producing behaviors that are guaranteed to asymptotically converge to an attractor is an important concept of traditional robot learning [6], [7].”
InfluentialTransferable Environment Model With Disentangled Dynamics
Yan, Q., Guo, S., Chen, D., Yang, Z., & Chen, F. (2019) · IEEE Access · doi:10.1109/access.2019.2927028 · Cited by 3
“[41] S. M. Khansari-Zadeh and A. Billard, ‘‘Learning stable nonlinear dynamical systems with Gaussian mixture models,’’ IEEE Trans.”
MethodAction-Conditional Recurrent Kalman Networks For Forward and Inverse Dynamics Learning
Shaj, V., Becker, P., Büchler, D., Pandya, H., van Duijkeren, N., Taylor, C. J., Hanheide, M., & Neumann, G. M. (2020) · Repository KITopen (Karlsruhe Institute of Technology) · doi:10.48550/arxiv.2010.10201 · Cited by 21
“Markov assumptions on the states and actions are applied to fit either the forward or inverse dynamics model to the training data. Authors used Linear Regression [6][7], Gaussian Mixture Regression [8][9], Gaussian Process Regression [10][11], Support Vector Regression [12], and feed forward neural networks [13]. Howev…”
Singh, S., Yang, L., & Ma, H. (2021) · IEEE World AI IoT Congress (AIIoT) · doi:10.1109/aiiot52608.2021.9454172 · Cited by 4
Xue, H., Michel, Y., & Lee, D. (2023) · Journal of Intelligent & Robotic Systems · doi:10.1007/s10846-023-02023-w · Cited by 3
“To the best of our knowledge, first order dynamical systems (DS) [17–21] have not been considered before in shared control for motion generation.”
BackgroundLearning search behaviour from humans
de Chambrier, G., & Billard, A. (2013) · IEEE International Conference on Robotics and Biomimetics (ROBIO) · doi:10.1109/robio.2013.6739521 · Cited by 3
“In many PbD research studies, single one-shot successful demonstrations (such as Pick & Place) in fully observable environments have been encoded through either statistical methods such as GMM with Stable Estimator of State Dynamics (SEDS) [6], in a latent space with Gaussian Process (GP) [7] or trajectory encoding me…”
MethodSelf-citationLearning externally modulated dynamical systems
Sommer, N., Kronander, K., & Billard, A. (2017) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2017.8206248 · Cited by 3
“Because stability is a major concern when dealing with DS, this has been addressed in our previous work [14] for a specific parametric form of DS, Gaussian Mixture Regression (GMR).”
BackgroundSelf-citationAugmenting Action Model Learning by Non-Geometric Features
Nematollahi, I., Kuhner, D., Welschehold, T., & Burgard, W. (2019) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2019.8794153 · Cited by 3
“present an approach that allows catching of object in flight using the dynamical system representation from [5].”
BackgroundMethodInfluentialCoupling of Arm Movements during Human-Robot Interaction: the handover case
Duarte, N. F., Raković, M., & Santos-Victor, J. (2019) · IEEE RO-MAN · doi:10.1109/ro-man46459.2019.8956352 · Cited by 3
“The DS encoding follows the joint distribution P(ξtn, ξ̇n) of [21] as Gaussian Mixture Models (GMM).”
MethodPhase-Synchronized Learning of Periodic Compliant Movement Primitives (P-CMPs)
Petrič, T. (2020) · Frontiers in Neurorobotics · doi:10.3389/fnbot.2020.599889 · Cited by 3
Vergara, C., Iregui, S., Schutter, J. D., & Aertbeliën, E. (2020) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros45743.2020.9340808 · Cited by 5
“For instance, in [17], authors shaped a reactive Lyapunov controller towards an attractor using demonstrations encoded by Gaussian Mixture Models (GMMs).”
MethodBa, H., Fan, J., Guo, X., & Hao, J. (2021) · International Conference on Advanced Robotics and Mechatronics (ICARM) · doi:10.1109/icarm52023.2021.9536131 · Cited by 6
Learning Motor Resonance in Human-Human and Human-Robot Interaction with Coupled Dynamical Systems
Duarte, N. F., Raković, M., & Santos-Victor, J. (2021) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48506.2021.9561378 · Cited by 5
“To compute the GMM parameters we use the stable estimator of dynamical systems (SEDS) approach [16], since it ensures assymptotic stability at the attractor point.”
BackgroundMethodInfluentialOn stability for learning human control strategy by demonstrations using SVM
Wang, Z., & Ou, Y. (2019) · Assembly Automation · doi:10.1108/aa-11-2018-0236 · Cited by 3
Learning Time-Invariant Reward Functions through Model-Based Inverse Reinforcement LearningPreprint
Davchev, T., Bechtle, S., Ramamoorthy, S., & Meier, F. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.03186 · Cited by 5
“BSfx: 1cGR: [17, 18, 19, 20, 21] learn stable, time-invariant dynamical systems by utilising a tem106 poral scaling variable similar to the one proposed in this work.”
Stochastic first-order learning for large-scale flexibly tied Gaussian mixture models
Pasande, M., Hosseini, R., & Araabi, B. N. (2024) · Pattern Recognition Letters · doi:10.1016/j.patrec.2023.12.021 · Cited by 2
Online model adaptation in Monte Carlo tree search planning
Zuccotto, M., Fusa, E., Castellini, A., & Farinelli, A. (2024) · Optimization and Engineering · doi:10.1007/s11081-024-09896-2 · Cited by 3
Learning Deep Dynamical Systems using Stable Neural ODEs
Sochopoulos, A., Gienger, M., & Vijayakumar, S. (2024) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10801826 · Cited by 4
“Furthermore, conventional DS learning approaches [6]–[8] necessitate access to state derivative information for the supervised learning of the underlying function f .”
InfluentialWang, R., Hu, W., & Ma, X. (2025) · Lecture notes in computer science · doi:10.1007/978-981-96-0774-7_21 · Cited by 2
Energy Approximated Dynamic Subattractor for Adjusting Obstacle Avoidance Trajectories
Dong, Y., Zeng, C., Jin, Z., Wang, N., & Yang, C. (2025) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2025.3591855 · Cited by 2
“It primarily employs two approaches: time-driven DS methods variants (e.g., DMP [4], ProMP [37], GPMP [38], tpDMP [39]) and state-driven DS methods, including GMM [5], SEDS [6], and LPV-DS [7].”
Zou, Y., Cheng, L., Yu, Z., Zhang, H., Han, L., Wang, Y., & Liu, Z. (2025) · Science China Technological Sciences · doi:10.1007/s11431-025-3029-1 · Cited by 2
“[21–23].”
Qiao, M., Qiu, Y., Li, A., Zhou, J. Y., & Deng, W. (2025) · Measurement Science and Technology · doi:10.1088/1361-6501/ae1314 · Cited by 3
“Assume the measurement noise η k follows a GMM [32] comprising M components, with its probability density function given by: where M is the number of Gaussian components, π i is the mixing coefficient of the i th Gaussian component satisfying 0 ⩽ π i ⩽ 1, and P denotes the Gaussian distribution…”
Arc-Length-Based Warping for Robot Skill Synthesis from Multiple Demonstrations
Braglia, G., Tebaldi, D., Lazzaretti, A. E., & Biagiotti, L. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11245947 · Cited by 11
“Readers interested in a detailed explanation of these metrics are encouraged to consult the additional material provided in this paper.”
One-Shot Robust Imitation Learning for Long-Horizon Visuomotor Tasks from Unsegmented Demonstrations
Wu, S., Wang, Y., & Huang, Y. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11246635 · Cited by 3
“One promising avenue to achieve perturbation resilience is to encode skills via an autonomous dynamical system (DS) [6], [7].”
An Alignment-Based Approach to Learning Motions From Demonstrations
Cuellar, A., Fourie, C. K., & Shah, J. (2025) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2025.3613956 · Cited by 2
“Existing methods can be partitioned into time-independent approaches, such as learned dynamical systems [2], [3], and time-dependent approaches, such as movement primitives [4], [5].”
Behavior-Controllable Stable Dynamics Models on Riemannian Configuration Manifolds
Lee, B., Lee, Y., Ha, J., & Park, F. C. (2025) · IEEE Transactions on Robotics · doi:10.1109/tro.2025.3647763 · Cited by 6
“For the Lyapunov-function-based methods, we consider stable estimator of dynamical systems (SEDSs) [2], which use predefined quadratic Lyapunov functions, stable deep dynamics models (SDDMs) [10], which rely on neural network Lyapunov functions, and LAGS-DSs , which introduce an adjustable…”
InfluentialAdaptive Stiffness Control for Series Elastic Actuators in Robotic Systems Using Dynamic Systems
Guo, K., Lu, Z., & Tan, Z. (2025) · Computational Intelligence · doi:10.1111/coin.70163 · Cited by 2
“A similar model, resem-blingStableEstimatorsofDynamicalSystems(SEDS)andLocally Active-Globally Stable Dynamical Systems (LAGS-DS), has been introduced in [20] and [21].”
Chen, R., Zhang, Z., Zhang, Z., Li, J., & Zhang, H. (2026) · Sensors · doi:10.3390/s26082387 · Cited by 2
Demonstrate once, execute on many: Kinematic intelligence for cross-robot skill transfer
Gupta, S., Salunkhe, D. H., & Billard, A. (2026) · Science Robotics · doi:10.1126/scirobotics.aea1995 · Cited by 4
Self-citationElectrostatic Gaussian Transistor for Real‐Time Probabilistic Inference
Han, Y., Yoo, Y., Kim, M. U., Jayasuriya, D., Ns, J., Subramanian, A., Jariwala, D., Kim, C., Trivedi, A. R., Kim, Y., et al. (2026) · Advanced Materials · doi:10.1002/adma.74242 · Cited by 2
Bautista-Ballester, J., Vergés–Llahí, J., & Puig, D. (2013) · Advances in intelligent systems and computing · doi:10.1007/978-3-319-03413-3_21 · Cited by 3
“Mapping Function Classification Low Level Robot Actions [23, 24, 29]”
BackgroundMovement Primitives as a Robotic Tool to Interpret Trajectories Through Learning-by-doing
Soltoggio, A., & Lemme, A. (2013) · Machine Intelligence Research · doi:10.1007/s11633-013-0734-9 · Cited by 3
“Keywords: Movement primitives, Learning, Pattern matching, Trajectory decomposition.”
Mohammad, Y. F. O., & Nishida, T. (2015) · IIAE International Conference on Industrial Application Engineering (ICIAE) · doi:10.12792/iciae2015.082 · Cited by 2
LSTM Learn Policy from Dynamical System of Demonstration Motions for Robot Imitation Learning
Yu, J., Yao, H., Zuo, G., & An, S. (2019) · IEEE International Conference on Cyber Technology in Automation, Control, and Intelligent Systems · doi:10.1109/cyber46603.2019.9066716 · Cited by 2
“Similar to [6], the constraints added here are as follows T ( ) 0 k A where k A and k B are defined according to Eq.”
BackgroundMethodInfluentialProbabilistic Adaptive Control for Robust Behavior Imitation
Jankowski, J., Girgin, H., & Calinon, S. (2021) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2021.3061310 · Cited by 2
Learning English Writing Skills from Images
Zhang, Y., Zou, Y., Li, H., Zhang, H., & Cheng, L. (2023) · IEEE International Conference on Development and Learning (ICDL) · doi:10.1109/icdl55364.2023.10364389 · Cited by 2
“[2]–[4].”
Imitation Learning-Based Visual Servoing for Tracking Moving Objects
Felici, R., Saveriano, M., Roveda, L., & Paolillo, A. (2024) · Springer proceedings in advanced robotics · doi:10.1007/978-3-031-55000-3_8 · Cited by 2
Learning Barrier-Certified Polynomial Dynamical Systems for Obstacle Avoidance with Robots
Schonger, M., Kussaba, H. T. M., Chen, L., Figueredo, L., Swikir, A., Billard, A., & Haddadin, S. (2024) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra57147.2024.10610828 · Cited by 2
“Encoding demonstrations in a globally stable dynamical system aims to accurately reproduce these for the robot to follow while ensuring that it will still converge towards the desired equilibrium point even if perturbations occur along the trajectory [3]–[5].”
BackgroundMethodInfluentialSelf-citationRobot Learning Method for Human-like Arm Skills Based on the Hybrid Primitive Framework
Li, J., Han, H., Hu, J., Lin, J., & Li, P. (2024) · Sensors · doi:10.3390/s24123964 · Cited by 3
“The algorithm has good adaptability to disturbances in time and space [21].”
A Distributed Neural Hybrid System Learning Framework in Modeling Complex Dynamical Systems
Yang, Y., Wang, T., & Xiang, W. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3417330 · Cited by 2
“Human handwriting demonstrations of (a) Spoon and (b) P shape from LASA data set [34], demonstrations contain seven trajectories, each of which has 1000 location samples.”
Coupled Conditional Neural Movement Primitives
Pekmezci, M., Uğur, E., & Öztop, E. (2024) · Neural Computing and Applications · doi:10.1007/s00521-024-10077-6 · Cited by 3
“…alternative to model the nonlinear term of the DMPs. Statistical modeling techniques, on the other hand, use stochastic models such as Gaussian mixture models [6, 43, 44] or temporal probabilistic models such as hidden Markov models[32, 33, 45] to encode the movement from several demonstrations.”
Robot Skill Generalization via Keypoint Integrated Soft Actor-Critic Gaussian Mixture Models
Nematollahi, I., Yankov, K., Burgard, W., & Welschehold, T. (2024) · Springer proceedings in advanced robotics · doi:10.1007/978-3-031-63596-0_15 · Cited by 3
Atsuta, H., Kawai, Y., & Asada, M. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10649930 · Cited by 2
“Instead of assuming specific equations, Khansari-Zadeh and Billard [5] proposed a general method in which a Gaussian mixture model is optimized to ensure convergence at trajectory endpoints.”
BackgroundMethodAdaptive Admittance Control with Dynamical Systems for Optimized Motion Adaptation
Guo, Y., Peng, G., & Yang, C. (2024) · International Conference on Advanced Robotics and Mechatronics (ICARM) · doi:10.1109/icarm62033.2024.10715781 · Cited by 2
“The basic idea of “motion learning [9]” is to encode trajectory for demonstrated tasks, which is not applicable for tasks that consider collision with the environment.”
BackgroundMethodInfluentialTransferring Shotcrete Skills to Robots
Wu, R., Gholami, S., Bonato, T., Munier, L., & Billard, A. (2024) · IEEE/RAS-EMBS International Conference on Biomedical Robotics and Biomechatronics (BioRob) · doi:10.1109/biorob60516.2024.10719911 · Cited by 2
“This choice is motivated by the method’s performance in handling nonlinear trajectories and its ability to exhibit local impedance-like behavior, which distinguishes it from other LfD methods like SEDS [15] and LPVDS [16].”
MethodSelf-citationLearning Multi-Reference Frame Skills from Demonstration with Task-Parameterized Gaussian Processes
Montero, M. R., Franzese, G., Kober, J., & Santina, C. D. (2024) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10803060 · Cited by 2
“4 illustrates the vector field of the learned dynamics with and without uncertainty minimization in reproducing the dynamics of drawing a letter “G” from the LASA dataset [3].”
A Trajectory Planning Method Based on Imitation Learning for Hydraulic Robotic Manipulators
Liu, X., Zheng, Y., Gao, S., Song, R., Gao, R., & Li, Y. (2024) · IEEE International Conference on Robotics and Biomimetics (ROBIO) · doi:10.1109/robio64047.2024.10907648 · Cited by 2
“Data modeling from trajectory demonstrations entails expressing the demonstrated trajectory data using mathematical models, such as probability statistical models [4] or dynamic system models [11].”
CDPMM-DMP: Conditional Dirichlet Process Mixture Model-Based Dynamic Movement Primitives
Jiang, H., He, J., & Duan, X. (2025) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2025.3559696 · Cited by 4
“Therefore, when learning from a single demonstration, DMP o ff ers a robust and accurate way to encode complex dynamics [33].”
Robust Dynamical Systems Learning for Robot-Assisted Point-to-Point Rehabilitation Training
Zhang, H., & Cheng, L. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-6960-8_17 · Cited by 2
Generalizable Motion Policies Through Keypoint Parameterization and Transportation Maps
Franzese, G., Prakash, R., Santina, C. D., & Kober, J. (2025) · IEEE Transactions on Robotics · doi:10.1109/tro.2025.3582821 · Cited by 9
“The proposed algorithm will “transport” the original labels on the new task configuration where a new policy can be re-fitted, making the generalization algorithm agnostic to the methodusedforpolicylearning,e.g.,DMP[6],KMP[7],GP[4], NN[8],GMM[9],LPV[10],etc.”
Moghani, M., Nelson, N., Ghanem, M., Diaz‐Pinto, A., Hari, K., Azizian, M., Goldberg, K., Huver, S. D., & Garg, A. (2025) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra55743.2025.11127797 · Cited by 3
“This approach has proven effective for various robotics applications [11]–[13] and manipulation tasks [14]–[16], where replicating expert behavior is crucial.”
Negative Imaginary Neural ODEs: Learning to Control Mechanical Systems with Stability Guarantees
Shi, K., Wang, R., & Manchester, I. R. (2025) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2025.11.060 · Cited by 3
Li, T., Sun, S., Aditya, S. S., & Figueroa, N. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11246142 · Cited by 12
“The proposed convex optimization for P-QLF learning is evaluated on the LASA datasets [24], a collection of 30 human handwriting motions in 2D, each comprising 7 trajectories, with a total of 7000 observations.”
Framework for Learning and Adaptation of Humanoid Robot Skills to Task Constraints
García, D. H., Monje, C. A., & Balaguer, C. (2013) · Advances in intelligent systems and computing · doi:10.1007/978-3-319-03413-3_41 · Cited by 2
“By being time-invariant and globally asymptotically stable at the target, the DS estimated with SEDS are able to respond immediately and appropriately to perturbations that could be encountered during reproduction of the motion [11].”
BackgroundMethodSampling Approach Matters: Active Learning for Robotic Language Acquisition
Pillai, N., Raff, E., Ferraro, F., & Matuszek, C. (2020) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata50022.2020.9378415 · Cited by 3
“We take advantage of that body of research to select our set of experimental approaches, which include sample selection via Gaussian mixture models [17], [31] and Determinantal Point Processes (DPPs) [38], which have proven effective in modeling diversity [21], [74].”
MethodOlanrewaju, R. O., Barry, T. S., Muse, A. H., & Habineza, A. (2021) · Mathematical theory and modeling · doi:10.7176/mtm/11-3-02 · Cited by 2
Tsinganos, K., Chatzilygeroudis, K. I., Hadjivelichkov, D., Komninos, T., Dermatas, E., & Kanoulas, D. (2022) · Frontiers in Robotics and AI · doi:10.3389/frobt.2022.974537 · Cited by 2
“Using this type of approaches, robotic manipulators can learn to draw digits or perform complex trajectories using only a few demonstrations (Khansari-Zadeh and Billard, 2011), or even humanoids can learn how to navigate and co-manipulate objects (Figueroa et al.”
BackgroundMethodLearning to Play Mini-Golf from Human Demonstration using Autonomous Dynamical Systems
Khansari-Zadeh, S. M., & Billard, A. (2011) · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · Cited by 2
Self-citationLahroodi, M., & Konukseven, E. İ. (2015) · The proceedings of the IEEE Conference on Control Applications/The proceedings of the ... IEEE Conference on Control Applications · doi:10.1109/cca.2015.7320626 · Cited by 2
“Of course this study starts with simple tasks like walking[8], reaching [9]and goes toward more complex tasks like running in rough terrain[10].”
BackgroundGeneration and adaptation of robot skills models
García, D. H., Monje, C. A., & Balaguer, C. (2014) · IEEE-RAS International Conference on Humanoid Robots (Humanoids) · doi:10.1109/humanoids.2014.7041356 · Cited by 2
“By being time-invariant and globally asymptotically stable at the target, the DS estimated with SEDS are able to respond immediately and appropriately to perturbations that could be encountered during reproduction of the motion ([4]).”
BackgroundMethodInfluentialHow Rich Motor Skills Empower Robots at Last: Insights and Progress of the AMARSi Project
Soltoggio, A., & Steil, J. J. (2012) · Künstliche Intell. · doi:10.1007/s13218-012-0192-5 · Cited by 2
“In AMARSi the application of learning to robotics has further advanced, in particular imitation learning [11, 19], learning of nonlinear systems [17, 18] and learning of motion dynamics for catching [20].”
BackgroundAdaptation of Robot Skills Models to New Task Contraints
García, D. H., Monje, C. A., & Balaguer, C. (2015) · International Journal of Humanoid Robotics · doi:10.1142/s0219843615500243 · Cited by 2
Chen, J., & Lau, H. Y. K. (2016) · IEEE Workshop on Advanced Robotics and its Social Impacts (ARSO) · doi:10.1109/arso.2016.7736260 · Cited by 2
WMRA skill learning through segmentation of demonstration
Yao, Y., Chi, M., Liu, Y., Du, Q., & Zhaomin, W. (2017) · International Conference on Advanced Robotics and Mechatronics (ICARM) · doi:10.1109/icarm.2017.8273220 · Cited by 2
Smooth trajectory generation for soft catching a flying object with an aerial vehicle
Lee, S., Seo, H., Choi, S., Kim, H., & Kim, H. J. (2017) · Asian Control Conference (ASCC) · doi:10.1109/ascc.2017.8287271 · Cited by 2
“Our approach is based on the SEDS which guarantees global asymptotic stability at the target [8].”
MethodExtended Reproduction of Demonstration Motion Using Variational Autoencoder
Takahashi, D., & Katsura, S. (2018) · IEEE International Symposium on Industrial Electronics (ISIE) · doi:10.1109/isie.2018.8433683 · Cited by 2
“As a method for this purpose, there are methods of constructing a probabilistic model based on demonstrations [9], [10], methods combining reinforcement learning and neural networks [11]–[13], a method of recurrent neural network [14], and a method of generating a potential field [15].”
MethodDutta, S., Kumar, S., & Behera, L. (2018) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2018.8489055 · Cited by 2
Robot Motor Skill Acquisition with Learning in Two Spaces
Fu, J. Y., Cao, C., Du, J., & Shen, S. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-27529-7_33 · Cited by 2
Improved Learning Accuracy for Learning Stable Control from Human Demonstrations
Jin, S., Wang, Z., Ou, Y., & Zhou, Y. (2019) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros40897.2019.8968274 · Cited by 2
“By learning a Stable Estimator of Dynamical System, a new framework was provided for the robot to learn point-to-point motions [11].”
BackgroundMethodInfluentialAdaptive Collision-Free Reaching Skill Learning from Demonstration
Ogenyi, U. E., Zhou, D., Ju, Z., & Liu, H. (2020) · Communications in computer and information science · doi:10.1007/978-981-33-4932-2_8 · Cited by 2
Trajectory Modeling by Distributed Gaussian Processes in Multiagent Systems
Xin, D.-J., & Shi, L. (2022) · Sensors · doi:10.3390/s22207887 · Cited by 2
“[5].”
BackgroundUsing Behavioral Primitives to Model Students’ Digital Behavior
Dermy, O., Roussanaly, A., & Boyer, A. E. (2022) · Procedia Computer Science · doi:10.1016/j.procs.2022.09.302 · Cited by 5
Integrated Bi-Manual Motion Generation and Control shaped for Probabilistic Movement Primitives
Vorndamme, J., Carvalho, J., Laha, R., Koert, D., Figueredo, L., Peters, J., & Haddadin, S. (2022) · IEEE-RAS International Conference on Humanoid Robots (Humanoids) · doi:10.1109/humanoids53995.2022.10000149 · Cited by 4
“Probabilistic Movement Primitives (ProMPs) [1] and Stable Estimator of Dynamical Systems (SEDS) [4], enable to capture the inherent stochasticity within human’s demonstrations.”
BackgroundHuman Intention Understanding and Trajectory Planning Based on Multi-modal Data
Liu, C., Cao, X., & Li, X. (2023) · Communications in computer and information science · doi:10.1007/978-981-99-0617-8_26 · Cited by 3
Dynamical System-based Imitation Learning for Visual Servoing using the Large Projection Formulation
Paolillo, A., Giordano, P. R., & Saveriano, M. (2023) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48891.2023.10160935 · Cited by 8
Learning Stable Nonlinear Dynamical System from One Demonstration
Zhang, Y., Han, L.-J., Wang, Z., Xia, X., Li, H., & Cheng, L. (2023) · Lecture notes in computer science · doi:10.1007/978-981-99-8070-3_36 · Cited by 2
Mahjoub, R. (2024) · Heliyon · doi:10.1016/j.heliyon.2024.e33605 · Cited by 2
“Moreover, these approaches like Gaussian mixture models can stabilize the dynamical system at the attractor point in trajectory space but others like expectation-maximization cannot make sure for guiding the dynamical system to target or encode the span of data [8,14].”
Structural Risk Minimization for Learning Nonlinear Dynamics
Stamouli, C. J., Chatzipantazis, E., & Pappas, G. J. (2024) · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · doi:10.23919/acc60939.2024.10644471 · Cited by 2
Imitation learning framework based on principal component analysis
Park, G., & Konno, A. (2015) · Advanced Robotics · doi:10.1080/01691864.2015.1007084 · Cited by 3
“[29,30] DMPbasically requires areferencetrajectorygeneratedusingGMR,weightedleast-square (WLS), etc. Khansari-Zadeh and Billard presented the stable estimator of dynamical systems (SEDS) to learn arbitrary discrete motions by modeling them as a nonlinear autonomous DS.[31] DS, DMP, and SEDS are designed to handle the…”
BackgroundLearning Replanning Policies With Direct Policy Search
Brandherm, F., Peters, J., Neumann, G. M., & Akrour, R. (2019) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2019.2901656 · Cited by 5
“Khansari-Zadeh and Billard introduced autonomous dynamic systems [25] which is an alternative formulation of dynamical trajectories that models the dynamical system with Gaussian mixture models and has the ability to adapt to perturbations instantly.”
BackgroundCombining decision making and dynamical systems for monitoring and executing manipulation tasks
Saveriano, M., & Piater, J. (2020) · e+i Elektrotechnik und Informationstechnik · doi:10.1007/s00502-020-00816-7 · Cited by 3
Learning Optimized Human Motion via Phase Space Analysis
Gesel, P., Mikulis-Borsoi, F. A. S., LaRoche, D. P., Arthanat, S., & Begum, M. (2020) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros45743.2020.9340773 · Cited by 2
“State of the art methods for human motion learning include dynamic system learning [12], [13], inverse optimal control (IOC) [14], [15], and statistical learning [16], [17], approaches.”
BackgroundSultangazin, A., Fraile, L., & Tabuada, P. (2021) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc45484.2021.9683193 · Cited by 2
Learning Motor Resonance in Human-Human and Human-Robot Interaction with Coupled Dynamical SystemPreprint
Duarte, N. F., Raković, M., & Santos-Victor, J. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1905.04072 · Cited by 2
“To compute the GMM parameters we use the stable estimator of dynamical systems (SEDS) approach [16], since it ensures assymptotic stability at the attractor point.”
BackgroundMethodInfluentialImitation Learning of Robot Policies by Combining Language, Vision and DemonstrationPreprint
Stepputtis, S., Campbell, J., Phielipp, M., Baral, C. R., & Amor, H. B. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1911.11744 · Cited by 5
“…part of the network, which is a DMP [Schaal, 1999], allowing us leverage a large body of research regarding their behavior and stability, while also allowing other extensions of DMPs [Amor et al., 2014, Paraschos et al., 2013, Khansari-Zadeh and Billard, 2011] to be incorporated to our framework.”
Generation of New Robot Skills from Learned Skills
García, D. H., Monje, C. A., & Balaguer, C. (2014) · IFAC Proceedings Volumes · doi:10.3182/20140824-6-za-1003.02232 · Cited by 1
A novel model for robots to avoid obstacles based on tensor analysis and differential geometry
Jin, S., Ou, Y., Wu, X., & Feng, W. (2016) · IEEE International Conference on Real-time Computing and Robotics (RCAR) · doi:10.1109/rcar.2016.7784024 · Cited by 1
Dynamical Systems based Obstacle Avoidance with Workspace Constraint for Manipulators
Zheng, D., Wu, X., & Pang, J. (2021) · Chinese Control and Decision Conference (CCDC) · doi:10.1109/ccdc52312.2021.9601889 · Cited by 1
A Novel Approach for Efficient Gaussian Mixture Model using Dynamics-motivated Optimal Excitation
Kim, T., Jeong, J., Kong, T., Lee, H., & Oh, S. (2024) · IEEE International Symposium on Industrial Electronics (ISIE) · doi:10.1109/isie54533.2024.10595794 · Cited by 1
“Yet, some methods treat models as black boxes and use random data for training, which can invalidate results and extend training time [10].”
MethodAdaptive Stiffness Control of Series Elastic Actuator Manipulators Based on Dynamic System
Guo, K., Zeng, C., Si, W., Wang, N., & Yang, C. (2024) · IEEE International Conference on Automation Science and Engineering (CASE) · doi:10.1109/case59546.2024.10711391 · Cited by 1
“A similar model, resembling Stable Estimators of Dynamical Systems (SEDS) and Locally Active-Globally Stable Dynamical Systems (LAGS-DS), has been introduced in references [17] and [18].”
MethodComparison of motion encoding frameworks on human manipulation actions
Jahn, L., Wörgötter, F., & Kulvičius, T. (2024) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2024.104869 · Cited by 1
“There are multiple ways to encode trajectories using GMMs [9], [13], [7], [8].”
InfluentialYang, Y., Wang, T., & Xiang, W. (2025) · Neural Networks · doi:10.1016/j.neunet.2025.107261 · Cited by 2
Metadata
- Authors
- Khansari-Zadeh, S. M., & Billard, A.
- Affiliations
- École Polytechnique Fédérale de Lausanne (CH)
- ORCID
- Billard, Aude
- Venue
- IEEE Transactions on Robotics
- Published
- 2011-10
- Volume
- 27
- Issue
- 5
- Pages
- 943-957
- ISSN
- 1552-3098 (ISSN-L)
- 1941-0468
- OpenAlex
- W2129202194
- Cited by
- 771
- References
- 43
- Citations by year
- 2026: 41 · 2025: 71 · 2024: 73 · 2023: 74 · 2022: 47 · 2021: 67 · 2020: 65 · 2019: 69 · 2018: 52 · 2017: 43 · 2016: 31 · 2015: 42 · 2014: 41 · 2013: 28 · 2012: 22
- FWCI
- 27.98
- Type
- journal article
- Language
- en
- Open access
- green
- infoscience.epfl.ch
- Locations
- IEEE Transactions on Robotics (publishedVersion)
- Infoscience (Ecole Polytechnique Fédérale de Lausanne) (submittedVersion)
- Infoscience (Ecole Polytechnique Fédérale de Lausanne) (submittedVersion, cc-by-nc-nd)
- Funders
- European Commission (FP7 ICT, 248311)
- FP7 Information and Communication Technologies
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Engineering › Control and Systems Engineering › Robot Manipulation and Learning
- Life Sciences › Neuroscience › Cognitive Neuroscience › Motor Control and Adaptation
- Keywords
- learning dynamical systems, global asymptotic stability, Gaussian mixture model, time-invariant systems, robustness to disturbances
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.