Takamitsu Matsubara - Profile on Academia.edu (original) (raw)

Takamitsu Matsubara

Tekin  MeriƧli related author profile picture

Matthew Ratcliffe related author profile picture

Adam M Croom related author profile picture

John Sutton related author profile picture

Saygin Abdikan related author profile picture

Fabio Cuzzolin related author profile picture

Gunnar Declerck related author profile picture

Fusun Balik Sanli related author profile picture

Stefano Marco Maria De Rossi related author profile picture

Tommaso  Ranzani related author profile picture

Uploads

Papers by Takamitsu Matsubara

Research paper thumbnail of Real-time stylistic prediction for whole-body human motions

Neural Networks, 2012

The ability to predict human motion is crucial in several contexts such as human tracking by comp... more The ability to predict human motion is crucial in several contexts such as human tracking by computer vision and the synthesis of human-like computer graphics. Previous work has focused on off-line processes with well-segmented data; however, many applications such as robotics require real-time control with efficient computation. In this paper, we propose a novel approach called real-time stylistic prediction for whole-body human motions to satisfy these requirements. This approach uses a novel generative model to represent a whole-body human motion including rhythmic motion (e.g., walking) and discrete motion (e.g., jumping). The generative model is composed of a low-dimensional state (phase) dynamics and a two-factor observation model, allowing it to capture the diversity of motion styles in humans. A real-time adaptation algorithm was derived to estimate both state variables and style parameter of the model from non-stationary unlabeled sequential observations. Moreover, with a simple modification, the algorithm allows real-time adaptation even from incomplete (partial) observations. Based on the estimated state and style, a future motion sequence can be accurately predicted. In our implementation, it takes less than 15ms for both adaptation and prediction at each observation. Our realtime stylistic prediction was evaluated for human walking, running, and jumping behaviors.

Research paper thumbnail of Task-adaptive inertial parameter estimation of rigid-body dynamics with modeling error for model-based control using covariate shift adaptation

Task-adaptive inertial parameter estimation of rigid-body dynamics with modeling error for model-based control using covariate shift adaptation

2014 IEEE/ASME International Conference on Advanced Intelligent Mechatronics, 2014

In this paper we consider the inertial parameter estimation problem of rigid-body dynamics with m... more In this paper we consider the inertial parameter estimation problem of rigid-body dynamics with modeling errors for model-based control. Our approach focuses on the task-specific subspace, i.e., it estimates the task-adaptive inertial parameter set to be more suitable for accurately performing a given task. We present a task-adaptive inertial parameter estimation procedure using a modern statistical supervised learning framework called covariate shift adaptation equipped with a direct importance estimation method. The effectiveness of the proposed method is investigated on the trajectory tracking task with an anthropomorphic manipulator model in simulations.

Research paper thumbnail of Real-time stylistic prediction for whole-body human motions

Neural Networks, 2012

The ability to predict human motion is crucial in several contexts such as human tracking by comp... more The ability to predict human motion is crucial in several contexts such as human tracking by computer vision and the synthesis of human-like computer graphics. Previous work has focused on off-line processes with well-segmented data; however, many applications such as robotics require real-time control with efficient computation. In this paper, we propose a novel approach called real-time stylistic prediction for whole-body human motions to satisfy these requirements. This approach uses a novel generative model to represent a whole-body human motion including rhythmic motion (e.g., walking) and discrete motion (e.g., jumping). The generative model is composed of a low-dimensional state (phase) dynamics and a two-factor observation model, allowing it to capture the diversity of motion styles in humans. A real-time adaptation algorithm was derived to estimate both state variables and style parameter of the model from non-stationary unlabeled sequential observations. Moreover, with a simple modification, the algorithm allows real-time adaptation even from incomplete (partial) observations. Based on the estimated state and style, a future motion sequence can be accurately predicted. In our implementation, it takes less than 15ms for both adaptation and prediction at each observation. Our realtime stylistic prediction was evaluated for human walking, running, and jumping behaviors.

Research paper thumbnail of Task-adaptive inertial parameter estimation of rigid-body dynamics with modeling error for model-based control using covariate shift adaptation

Task-adaptive inertial parameter estimation of rigid-body dynamics with modeling error for model-based control using covariate shift adaptation

2014 IEEE/ASME International Conference on Advanced Intelligent Mechatronics, 2014

In this paper we consider the inertial parameter estimation problem of rigid-body dynamics with m... more In this paper we consider the inertial parameter estimation problem of rigid-body dynamics with modeling errors for model-based control. Our approach focuses on the task-specific subspace, i.e., it estimates the task-adaptive inertial parameter set to be more suitable for accurately performing a given task. We present a task-adaptive inertial parameter estimation procedure using a modern statistical supervised learning framework called covariate shift adaptation equipped with a direct importance estimation method. The effectiveness of the proposed method is investigated on the trajectory tracking task with an anthropomorphic manipulator model in simulations.

Log In