Model learning for robot control: a survey

By Duy Nguyen-Tuong, Jan Peters. 2011

557 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Control and Systems Engineering, Reinforcement Learning, Robotics & Control

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Control and Systems Engineering

In Reinforcement Learning

In Robotics & Control

Citations in the corpus

Domain randomization for transferring deep neural networks from simulation to the real world (2017, background): “Even with strong system identification, the real world has unmodeled physical effects like nonrigidity, gear backlash, wear-and-tear, and fluid dynamics that are not captured by current physics simulators (though learning techniques may help bridge this gap [33]).”
Financial Time Series Forecasting with Deep Learning : A Systematic Literature Review: 2005-2019 (2019, methodology): “Model-based methods are mostly used in the application of robotics, and control algorithms [79].”
Active learning of inverse models with intrinsically motivated goal exploration in robots (2013, neutral): “Advanced statistical learning techniques dedicated to incremental high-dimensional regression have been elaborated recently, such as [1,2].”
DeepMPC: Learning Deep Latent Features for Model Predictive Control (2015, methodology): “Model learning for robot control has also been a very active area, and we refer the reader to a review of work in the area by Nguyen-Tuong and Peters [31].”
An Algorithmic Perspective on Imitation Learning (2018, methodology): “As indicated by [Nguyen-Tuong and Peters, 2011], various methods have been developed for model learning.”
Control Strategies for Soft Robotic Manipulators: A Survey (2018, unknown): “In addition, machine learning algorithms incorporating prior knowledge of the system would also provide a way for faster and more stable learning.(73) Another overlooked topic is the importance of the low-level controllers (actuator dynamics) in the overall stability and response of the higher level”
Identifying Causal Structure in Large-Scale Kinetic Systems (2018, background): “nvariant models, however, may transfer to other areas of applications, too. In robotics, for example, the concept of model learning and structure search is becoming increasingly more prominent [e.g., Nguyen-Tuong and Peters, 2011, Peters et al., 2016b]. Future extensions may also include the applica”
Black-box data-efficient policy search for robotics (2017, background): “One of the best ways to take advantage of this sequential state recording is to learn a dynamical model of the robot [11], and then exploit it either for model-predictive control [12] or to find an optimal policy offline [7].”
Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning (2021, background): “Moreover, adaptive control approaches tend to “overfit” to the latest observations and convergence to the true parameters is generally not guaranteed (23, 24).”
Trends and challenges in robot manipulation (2019, methodology): “Learning has been used to model contacts (80) and is also beneficial for reducing control dimensions by determining latent space, as required in bimanual dynamics (65).”
Deliberation for autonomous robots: A survey (2017, background): “• Sensory-motor control models, as in [47, 121, 120, 164], and most of the work surveyed in [197];”
On-line regression algorithms for learning mechanical models of robots: A survey (2011, methodology): “The control methods based on such models are not covered in the paper, we refer the reader to Nguyen-Tuong & Peters [71], Salaün et al.”
Reset-free Trial-and-Error Learning for Robot Damage Recovery (2016, methodology): “, when using Gaussian processes for models [22]).”
A Physics-Based Model Prior for Object-Oriented MDPs (2014, methodology): “…object the function f(st, at)→ s′t can be written as: f1(st, at) + . . . fn(st, at) + = s1t ′. . . snt ′ (3) For these individual predictors we use locally-weighted regression (we summarize the main properties of LWR here, but for a more thorough overview see (Nguyen-Tuong & Peters, 2011)).”
On-line policy learning and adaptation for real-time personalization of an artificial pancreas (2015, methodology): “Particularly, in the work of Nguyen-Tuong and Peters (2011a) an incremental online sparsiﬁcation (IOS) method has been proposed which is mainly inspired in the Kernel Recursive Least-Squares algorithm proposed by Engel et al. (2004).”
Reinforcement Learning in Continuous State and Action Spaces (2012, background): “For a recent survey on model-learning algorithms, see Nguyen-Tuong and Peters (2011).”
Deep Haptic Model Predictive Control for Robot-Assisted Dressing (2017, methodology): “Learning models with neural networks for robot control is common throughout many robotic control approaches [19].”