Tao Zhou

UCLA · Computer Science

We present a biomimetic framework for human neuromuscular and visuomotor control that promises to be of interest to researchers developing humanoid robots. Our framework features a biomechanically simulated human musculoskeletal model, actuated by numerous skeletal muscles, with realistic eyes driven by extraocular and intraocular muscles, whose op...

We present a novel, biomimetic model of the eye for realistic virtual human animation. We also introduce a deep learning approach to oculomotor control that is compatible with our biomechanical eye model. Our eye model consists of the following functional components: (i) submodels of the 6 extraocular muscles that actuate realistic eye movements, (...

We present a novel, biomimetic model of the eye for realistic virtual human animation. We also introduce a deep learning approach to oculomotor control that is compatible with our biomechanical eye model. Our eye model consists of the following functional components: (i) submodels of the 6 extraocular muscles that actuate realistic eye movements, (...

Doodling is a useful and common intelligent skill that people can learn and master. In this work, we propose a two-stage learning framework to teach a machine to doodle in a simulated painting environment via Stroke Demonstration and deep Q-learning (SDQ). The developed system, Doodle-SDQ, generates a sequence of pen actions to reproduce a referenc...

Doodling is a useful and common intelligent skill that people can learn and master. In this work, we propose a two-stage learning framework to teach a machine to doodle in a simulated painting environment via Stroke Demonstration and deep Q-learning (SDQ). The developed system, Doodle-SDQ, generates a sequence of pen actions to reproduce a referenc...

We introduce a biomimetic framework for human sensorimotor control, which features a biomechanically simulated human musculoskeletal model actuated by numerous muscles, with eyes whose retinas have nonuniformly distributed photoreceptors. The virtual human's sensorimotor control system comprises 20 trained deep neural networks (DNNs), half constitu...

Following the recent progress in image classification and captioning using deep learning, we develop a novel natural language person retrieval system based on an attention mechanism. More specifically, given the description of a person, the goal is to localize the person in an image. To this end, we first construct a benchmark dataset for natural l...

Following the recent progress in image classification and captioning using deep learning, we develop a novel natural language person retrieval system based on an attention mechanism. More specifically, given the description of a person, the goal is to localize the person in an image. To this end, we first construct a benchmark dataset for natural l...

Following the recent progress in image classification and image captioning using deep learning, we developed a novel person retrieval system using natural language, which to our knowledge is first of its kind. Our system employs a state-of-the-art deep learning based natural language object retrieval framework to detect and retrieve people in image...

equifinality and its violations in multi-joint positional tasks

We explored the phenomenon of unintentional changes in the apparent stiffness of the human arm produced by transient changes in the external force. The subjects performed a positional task against a constant baseline force and were instructed not to react to changes in the force. A HapticMaster robot produced a smooth force increase (a perturbation...

This study used the framework of the referent configuration hypothesis and slow changes in the external conditions during vertical oscillation of a hand-held object to infer the characteristics of hypothetical control variables. The study had two main objectives: (1) to show that hypothetical control variables, namely, referent coordinates and appa...

We tested predictions of a hierarchical scheme on the control of natural movements with referent body configurations. Subjects occupied an initial hand position against a bias force generated by a HapticMaster robot. A smooth force perturbation was applied to the hand consisting of an increase in the bias force, keeping it at a new level for 5 s, a...

The authors explored the phenomenon of unintentional changes in the equilibrium state of a multijoint effector produced by transient changes in the external force. The subjects performed a position-holding task against a constant force produced by a robot and were instructed not to intervene voluntarily with movements produced by changes in the rob...

We used the framework of the referent configuration (RC) hypothesis to reconstruct the hand and grip aperture referent trajectories during vertical oscillation of a hand-held object. Within the RC hypothesis, mechanical variables are consequences of shifts in referent values for salient body coordinates. For the studied task, this translates into t...

We tested several predictions of a recent theory that combines the ideas of control with referent configurations, hierarchical control, and the uncontrolled manifold (UCM) hypothesis. In particular, we tested a hypothesis that unintentional changes in hand coordinate can happen following a long-lasting transient perturbation. The subjects grasped a...

We used robot-generated perturbations applied during position-holding tasks to explore stability of induced unintentional movements in a multidimensional space of muscle activations. Healthy subjects held the handle of a robot against a constant bias force and were instructed not to interfere with hand movements produced by changes in the external...

We tested a prediction that equifinality at the task level may be accompanied by violations of equifinality within the redundant set of elemental variables. Seated subjects grasped a handle, and occupied an initial arm configuration against a bias force produced by a robot. The robot applied a smooth, transient change in the force (perturbation). T...

We studied a mechanism of feed-forward control of a multi-finger action, namely anticipatory synergy adjustments (ASAs), prior to a quick force correction in response to a change in the gain of the visual feedback. Synergies were defined as co-varied across trials adjustments of commands to fingers that stabilized (decreased variance of) the total...

In this paper, control of a nonholonomic mobile manipulators with unknown dynamics is considered using Diagonal Recurrent Neural Network. The system is subject to nonholonomic constraint. The adaptive recurrent neural network controller is presented to deal with the unmodelled dynamics in the system. The proposed control strategy guarantees that th...

In this article, motion/force control problem of a class of constrained mobile manipulators with unknown dynamics is considered. The system is subject to both holonomic and nonholonomic constraints. An adaptive recurrent neural network controller is proposed to deal with the unmodelled system dynamics. The proposed control strategy guarantees that...