Reconstructing State-Space from Movie Using Convolutional Autoencoder for Robot Control

Takahara, Kazuma; Ikemoto, Shuhei; Hosoda, Koh

doi:10.1007/978-3-030-01370-7_38

Cited by 1 publication

(2 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As the validation errors showed, over-fitting was not observed are difficult to configure to obtain the desired behavior of the manipulator. To address this problem, the method used for finding the latent feature from images, namely the methods proposed in [38,45], would be required to include additional cost functions for training a convolutional autoencoder that remains the intuitive meaning of distances in the latent space. Conversely, this would constitute an important aspect of the future work of this study, which will be discussed in the next section.…”

Section: Control By the Proposed Methodsmentioning

confidence: 99%

“…These images were used for training a convolutional autoencoder to derive a two-dimensional latent space where the posture of the robot is uniquely expressed and the dynamics can be modeled. To this end, we used a method that trains a convolutional autoencoder with a regulation cost based on a simultaneous forward modeling error of an NN (see [38,45] for the details). By using the trained convolutional autoencoder, encoding and decoding between images and two-dimensional vectors were made possible.…”

Section: Forward Model Learningmentioning

confidence: 99%

See 1 more Smart Citation

Neural Model Extraction for Model-Based Control of a Neural Network Forward Model

et al. 2021

Self Cite

View full text Add to dashboard Cite

Neural networks have been widely used to model nonlinear systems that are difficult to formulate. Thus far, because neural networks are a radically different approach to mathematical modeling, control theory has not been applied to them, even if they approximate the nonlinear state equation of a control object. In this research, we propose a new approach-i.e., neural model extraction, that enables model-based control for a feed-forward neural network trained for a nonlinear state equation. Specifically, we propose a method for extracting the linear state equations that are equivalent to the neural network corresponding to given input vectors. We conducted simple simulations of a two degrees-of-freedom planar manipulator to verify how the proposed method enables model-based control on neural network forward models. Through simulations, where different settings of the manipulator's state observation are assumed, we successfully confirm the validity of the proposed method.

show abstract

Section: Control By the Proposed Methodsmentioning

confidence: 99%

Section: Forward Model Learningmentioning

confidence: 99%