Dynamic Movement Primitives -A Framework for Motor Control in Humans and Humanoid Robotics

Schaal, Stefan

doi:10.1007/4-431-31381-8_23

Cited by 434 publications

(232 citation statements)

References 47 publications

(28 reference statements)

Supporting

Mentioning

217

Contrasting

Unclassified

Order By: Relevance

“…For a policy parameter θ ∈ Θ (e.g. θ can be the gains of a PD-controller or the paramters of a motion primitive such as [9]), r(x, θ) gives the reward of choosing parameter θ under context x. The goal of PS is to find a stochastic policy π(θ|x) that maximizes the policy return…”

Section: A Problem Formulation and Notationmentioning

confidence: 99%

“…The parameter η is the Lagrangian multiplier to the KLdivergence constraint (8) and the parameter ω is the Lagrangian multiplier to the entropy constraint (9). The Lagrangian multipliers of the last constraints (10) cancel out.…”

Section: Learning On the Gating Policy Layermentioning

confidence: 99%

“…To generate a trajectory, we use Dynamic Movement Primitives (DMP) [9]. We use one DMP for each joint.…”

Section: A Experiments Setup Reaching Point Taskmentioning

confidence: 99%

See 2 more Smart Citations

Layered direct policy search for learning hierarchical skills

End

Akrour

Peters

et al. 2017

2017 IEEE International Conference on Robotics and Automation (ICRA)

View full text Add to dashboard Cite

Abstract-Solutions to real world robotic tasks often require complex behaviors in high dimensional continuous state and action spaces. Reinforcement Learning (RL) is aimed at learning such behaviors but often fails for lack of scalability. To address this issue, Hierarchical RL (HRL) algorithms leverage hierarchical policies to exploit the structure of a task. However, many HRL algorithms rely on task specific knowledge such as a set of predefined sub-policies or sub-goals. In this paper we propose a new HRL algorithm based on information theoretic principles to autonomously uncover a diverse set of sub-policies and their activation policies. Moreover, the learning process mirrors the policys structure and is thus also hierarchical, consisting of a set of independent optimization problems. The hierarchical structure of the learning process allows us to control the learning rate of the sub-policies and the gating individually and add specific information theoretic constraints to each layer to ensure the diversification of the subpolicies. We evaluate our algorithm on two high dimensional continuous tasks and experimentally demonstrate its ability to autonomously discover a rich set of sub-policies.

show abstract

Section: A Problem Formulation and Notationmentioning

confidence: 99%

Section: Learning On the Gating Policy Layermentioning

confidence: 99%

See 1 more Smart Citation

Layered direct policy search for learning hierarchical skills

End

Akrour

Peters

et al. 2017

2017 IEEE International Conference on Robotics and Automation (ICRA)

View full text Add to dashboard Cite

show abstract

“…The idea of modular control architecture has drawn much attention in the study of motor control in human and humanoid robots (Kawato, Furukawa et al 1987;Kawato 1999;Doya, Kimura et al 2001;Schaal 2003). It is proposed that instead of a single gigantic model with a large number of adjustable parameters, multiple smaller scaled models co-exist in the brain with each being responsible for only a small subset of actions.…”

Section: Modular Internal Models As Multi-model Adaptive Controlmentioning

confidence: 99%

“…Such a scheme has been applied to humanoid robots (Schaal 2003;Schaal, Peters et al 2004). The movement of the robot is divided into different dynamic movement primitives (DMP), which are basic units of action in the form of stable nonlinear attractors.…”

Section: Modular Internal Models As Multi-model Adaptive Controlmentioning

confidence: 99%

Internal models in sensorimotor integration: perspectives from adaptive control theory

2005

View full text Add to dashboard Cite

Internal model and adaptive control are empirical and mathematical paradigms that have evolved separately to describe learning control processes in brain systems and engineering systems, respectively. This paper presents a comprehensive appraisal of the correlation between these paradigms with a view to forging a unified theoretical framework that may benefit both disciplines. It is suggested that the classic equilibrium-point theory of impedance control of arm movement is analogous to continuous gain-scheduling or high-gain adaptive control within or across movement trials, respectively, and that the recently proposed inverse internal model is akin to adaptive sliding control originally for robotic manipulator applications. Modular internal models architecture for multiple motor tasks is a form of multi-model adaptive control. Stochastic methods such as generalized predictive control, reinforcement learning, Bayesian learning and Hebbian feedback covariance learning are reviewed and their possible relevance to motor control is discussed. Possible applicability of Luenberger observer and extended Kalman filter to state estimation problems such as sensorimotor prediction or the resolution of vestibular sensory ambiguity is also discussed. The important role played by vestibular system identification in postural control suggests an indirect adaptive control scheme whereby system states or parameters are explicitly estimated prior to the implementation of control. This interdisciplinary framework should facilitate the experimental elucidation of the mechanisms of internal model in sensorimotor systems and the reverse engineering of such neural mechanisms into novel brain-inspired adaptive control paradigms in future.

show abstract

Conventional and non conventional body motions modelling and control. Application to the handwriting process

Abdelkrim

Benrejeb

2019

Asian Journal of Control

View full text Add to dashboard Cite

The handwriting process is one of the most important and complex activities of the human body. In recent years, modelling and analyzing and controlling it have received great attention leading to large important motion studies results. This paper attempts to present an overview of recent works about our proposed approaches and other ones to model and control this process. Hence, an overall model of the handwriting process has been proposed and developed by the coauthors, showing the interest of using the soft computing concepts, namely neural networks and fuzzy logic. The design of this model requires several undertaken steps. Starting from an experimental approach to acquire electromyographic signals and writing positions, different conventional and non conventional models have been proposed. Besides some control strategies have been considered in particular neural and fuzzy approaches in addition to non conventional neural and neuro-fuzzy Internal Model Control strategy. The proposed global model, inspired from the biological structure system, has been designed and validated through simulation results.

show abstract

Dynamic Movement Primitives -A Framework for Motor Control in Humans and Humanoid Robotics

Abstract: Given the continuous stream of movements that biological systems exhibit in their daily activities, an account for such versatility and creativity has to assume that movement sequences consist of segments, exe-

Cited by 434 publications

References 47 publications

Layered direct policy search for learning hierarchical skills

Layered direct policy search for learning hierarchical skills

Internal models in sensorimotor integration: perspectives from adaptive control theory

Conventional and non conventional body motions modelling and control. Application to the handwriting process

Contact Info

Product

Resources

About