Variational Bayesian learning of nonlinear hidden state-space models for model predictive control

By Tapani Raiko, Matti Tornio. 2009

43 citations recorded by the source corpus.

Subfields: Control and Systems Engineering, Probabilistic & Bayesian Methods, 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 Control and Systems Engineering

In Probabilistic & Bayesian Methods

In Robotics & Control

Citations in the corpus

Value Prediction Network (2017, methodology): “For continuous control problems, deep learning has been combined with model predictive control (MPC) [6, 19], a specific way of using an observation-prediction model.”
Weight Agnostic Neural Networks (2019, background): “Unlike the simpler task, it cannot be solved with a linear controller [87, 106].”
Deep Kalman Filters (2015, methodology): “Raiko & Tornio (2009) explored ways of using linear approximations and non-linear dynamical factor analysis in order to overcome these difﬁculties.”
Structured Inference Networks for Nonlinear State Space Models (2016, background): “We use the model (Raiko and Tornio 2009).”
Gaussian Processes for Data-Efficient Learning in Robotics and Control (2015, neutral): “Dynamics models were only learned in [18], [45], using RBF networks and multilayered perceptrons, respectively.”
Data-Efficient Reinforcement Learning with Probabilistic Model Predictive Control (2017, background): “swing-up and balancing task cannot be solved using a linear model [38].”
Diff-DAC: Distributed Actor-Critic for Average Multitask Deep Reinforcement Learning (2017, background): “, the force with which the cart supports the pole is a nonlinear function of the system state) and nonlinear features are required [6, 19].”
Dynamical Variational Autoencoders: A Comprehensive Review (2020, methodology): “, (Haykin, 2004; Raiko and Tornio, 2009), but it has been recently investigated under the VAE angle in two papers by the same authors (Krishnan et al.”
Model-based imitation learning by probabilistic trajectory matching (2013, methodology): “For instance, model-free methods based on trajectory sampling [27], [20] or model-based RL algorithms that learn forward models of the robot, and, subsequently, use them for predictions [12], [5], [23], [11], are suitable.”
Probabilistic Differential Dynamic Programming (2014, methodology): “PILCO: The recently proposed PILCO [14] framework has demonstrated state-of-the-art learning efficiency compared with other methods such as [21][22].”
PILCO: A Model-Based and Data-Efficient Approach to Policy Search (2011, methodology): “A linear controller is not capable of doing this (Raiko & Tornio, 2009).”
A Practical and Conceptual Framework for Learning in Control (2010, background): “Note that a linear controller is incapable to swing the pendulum up and to balance it in the inverted position (Raiko and Tornio, 2009).”
Data-efficient neuroevolution with kernel-based surrogate models (2018, methodology): “This task is more difficult than benchmarks used in many evolutionary computation publications, such as polebalancing, and cannot be solved with a linear controller [19], requiring networks to grow beyond their initial minimal state.”
A Separation Principle for Control in the Age of Deep Learning (2017, methodology): “Notice, however, that Variational Bayesian methods can be used as a partial solution to the ﬁrst problem [24, 32], as long as the function class covers the underlying data distribution, which remains an open problem for natural images.”