During Lipreading Training With Sentence Stimuli, Feedback Controls Learning and Generalization to Audiovisual Speech in Noise

Purpose: This study investigated the effects of external feedback on perceptual learning of visual speech during lipreading training with sentence stimuli. The goal was to improve visual-only (VO) speech recognition and increase accuracy of audiovisual (AV) speech recognition in noise. The rationale was that spoken word recognition depends on the accuracy of sublexical (phonemic/phonetic) speech perception; effective feedback during training must support sublexical perceptual learning. Method: Normal-hearing (NH) adults were assigned to one of three types of feedback: Sentence feedback was the entire sentence printed after responding to the stimulus. Word feedback was the correct response words and perceptually near but incorrect response words. Consonant feedback was correct response words and consonants in incorrect but perceptually near response words. Six training sessions were given. Pre- and posttraining testing included an untrained control group. Test stimuli were disyllable nonsense words for forced-choice consonant identification, and isolated words and sentences for open-set identification. Words and sentences were VO, AV, and audio-only (AO) with the audio in speech-shaped noise. Results: Lipreading accuracy increased during training. Pre- and posttraining tests of consonant identification showed no improvement beyond test–retest increases obtained by untrained controls. Isolated word recognition with a talker not seen during training showed that the control group improved more than the sentence group. Tests of untrained sentences showed that the consonant group significantly improved in all of the stimulus conditions (VO, AO, and AV). Its mean words correct scores increased by 9.2 percentage points for VO, 3.4 percentage points for AO, and 9.8 percentage points for AV stimuli. Conclusions: Consonant feedback during training with sentences stimuli significantly increased perceptual learning. The training generalized to untrained VO, AO, and AV sentence stimuli. Lipreading training has potential to significantly improve adults' face-to-face communication in noisy settings in which the talker can be seen.

Design and control of 2-DoF joystick using MR-fluid rotary actuator

The purpose of this paper is to design a control algorithm for a 2-DoF rotary joystick model. Firstly, the structure of the joystick, which composes of two magneto-rheological fluid actuators (shorten MRFA) with optimal configuration coupled perpendicularly by the gimbal mechanism to generate the friction torque for each independent rotary movement, is introduced. The control strategy of the designed joystick is then suggested. Really, because of two independent rotary movements, it is necessary to design two corresponding controllers. Due to hysteresis and nonlinear dynamic characteristics of the MRFA, controllers based an accurate dynamic model are difficult to realize. Hence, to release this issue, the proposed controller (named self-turning fuzzy controllers-STFC) will be built through the fuzzy logic algorithm in which the parameters of controllers are learned and trained online by Levenberg-Marquardt training algorithm. Finally, an experimental apparatus will be constructed to assess the effectiveness of the force feedback controls. Herein, three experimental cases are performed to compare the control performance of open-loop and close-loop control method, where the former is done through relationship between the force at the knob and the current supplied to coil while the latter is realized based on the proposed controller and PID controller. The experimental results provide strongly the ability of the proposed controller, meaning that the STFC is robust and tracks well the desirable force with high accuracy compared with both the PID controller and the open-loop control method.

Microtubule-mitochondrial attachment determines cell division symmetry and polarity in fission yeast

Association with microtubules inhibits the fission of mitochondria in Schizosachharomyces pombe. Here we show that this attachment of mitochondria to microtubules is an important cell intrinsic factor in determining division symmetry as well as maintaining polarity. By comparing mutant cells that exhibited enhanced attachment and no attachment of mitochondria to microtubules (Dnm1Δ and Mmb1Δ respectively), we show that microtubules in these mutants displayed aberrant dynamics compared to wild-type cells, which resulted in errors in nuclear positioning. This translated to cell division asymmetry in a significant proportion of both Dnm1Δ and Mmb1Δ cells. So too, microtubule pivoting was enhanced in both mitochondrial mutants, resulting in a fraction of the cells in these populations displaying polarity defects. The asymmetric division in Dnm1Δ and Mmb1Δ cells resulted in unequal distribution of mitochondria, with the daughter cell that received more mitochondria growing faster than the other daughter. Taken together, we show the existence of homeostatic feedback controls between mitochondria and microtubules in fission yeast, which directly influence mitochondrial partitioning and thereby, cell growth.

Feedback Optimality Condition for Nonconvex Discrete Linear-Quadratic Optimal Control Problem

The paper analyzed a non-convex linear-quadratic optimization problem in a discrete dynamic system. We obtained necessary optimality condition with feedback controls which allow a descent of the functional cost. Such controls are generated by the quadratic majorant of the cost. In contrast to the discrete maximum principle, this condition does not require any convexity properties of the problem.

Planar Five-link Biped Robot Movement over a Stepped Surface

Modeling the anthropomorphic robot movement is of great interest to researchers all over the world. At the same time, the movement control of a walking mechanism is always a high dimension challenge. The difficulty with the anthropomorphic robot control is also caused by the fact that such a mechanism has always a hybrid dynamics and represents a sequential change of two phases – the single support phase and the double support phase (phase of changing robot’s leg). At the single support phase and at another phase the behavior of the biped robot is described by a system of ordinary differential equations and by a system of linear algebraic equations, respectively.The task of biped robot movement control has been studied in detail for the case when the robot moves over the horizontal surface. Obstacles make the task significantly complicated. The paper considers the movement control of the biped robot over the surface that is a periodic alternation of horizontal sections and obstacles. The obstacles represent steps of the same height known. It is assumed that the lengths of horizontal sections and steps are known as well. The objective is to create a control that provides robot’s periodic movement over the specified surface according to inherent characteristics of a walking human.For the single support phase, the outputs are proposed, the equality of which to zero corresponds to the robot’s movement with a given set of characteristics. The paper presents the feedback controls that stabilize the proposed outputs for a finite amount of time. By choosing the feedback parameters, it is possible to adjust the stabilization time so that the outputs become equal to zero when reached the end of each step.It is shown that for the chosen control law, the problem of constructing the control of robot’s periodic movement is reduced to the solution of a nonlinear equation. In the paper, we discuss the approaches to solving this equation and present the results of numerical simulation.The results obtained can be used to solve the problem of providing control of the biped robot movement over the surfaces with obstacles of a more complicated shape.Modeling the anthropomorphic robot movement is of great interest to researchers all over the world. At the same time, the movement control of a walking mechanism is always a high dimension challenge. The difficulty with the anthropomorphic robot control is also caused by the fact that such a mechanism has always a hybrid dynamics and represents a sequential change of two phases – the single support phase and the double support phase (phase of changing robot’s leg). At the single support phase and at another phase the behavior of the biped robot is described by a system of ordinary differential equations and by a system of linear algebraic equations, respectively.The task of biped robot movement control has been studied in detail for the case when the robot moves over the horizontal surface. Obstacles make the task significantly complicated. The paper considers the movement control of the biped robot over the surface that is a periodic alternation of horizontal sections and obstacles. The obstacles represent steps of the same height known. It is assumed that the lengths of horizontal sections and steps are known as well. The objective is to create a control that provides robot’s periodic movement over the specified surface according to inherent characteristics of a walking human.For the single support phase, the outputs are proposed, the equality of which to zero corresponds to the robot’s movement with a given set of characteristics. The paper presents the feedback controls that stabilize the proposed outputs for a finite amount of time. By choosing the feedback parameters, it is possible to adjust the stabilization time so that the outputs become equal to zero when reached the end of each step.It is shown that for the chosen control law, the problem of constructing the control of robot’s periodic movement is reduced to the solution of a nonlinear equation. In the paper, we discuss the approaches to solving this equation and present the results of numerical simulation.The results obtained can be used to solve the problem of providing control of the biped robot movement over the surfaces with obstacles of a more complicated shape.

Bio Scaffolds

Abstract‘Bio Scaffolds’ explores a series of design tectonics that emerge from a co-creation between human, machine and natural intelligences. This research establishes an integral connection between form and materiality by enabling biological materials to become a co-creator within the design and fabrication process. In this research paper, we explore a hybrid between architectural aesthetics and biological agency by choreographing natural growth through form. ‘Bio Scaffolds’ explores a series of 3D printed biodegradable scaffolds that orchestrate both Mycelia growth and degradation through form. A robotic arm is introduced into the system that can respond to the organism’s natural behavior by injecting additional Mycelium culture into a series of sacrificial frameworks. Equipped with computer vision systems, feedback controls, scanning processes and a multi-functional end-effector, the machine tends to nature by reacting to its patterns of growth, moisture, and color variation. Using this cybernetic intelligence, developed between human, machine, and Mycelium, our intention is to generate unexpected structural and morphological forms that are represented via a series of 3D printed Mycelium enclosures. ‘Bio Scaffolds’ explores an interplay between biological and computational complexity through non anthropocentric micro habitats.

The Adaptive Dynamic Programming Toolbox

The paper develops the adaptive dynamic programming toolbox (ADPT), which is a MATLAB-based software package and computationally solves optimal control problems for continuous-time control-affine systems. The ADPT produces approximate optimal feedback controls by employing the adaptive dynamic programming technique and solving the Hamilton–Jacobi–Bellman equation approximately. A novel implementation method is derived to optimize the memory consumption by the ADPT throughout its execution. The ADPT supports two working modes: model-based mode and model-free mode. In the former mode, the ADPT computes optimal feedback controls provided the system dynamics. In the latter mode, optimal feedback controls are generated from the measurements of system trajectories, without the requirement of knowledge of the system model. Multiple setting options are provided in the ADPT, such that various customized circumstances can be accommodated. Compared to other popular software toolboxes for optimal control, the ADPT features computational precision and time efficiency, which is illustrated with its applications to a highly non-linear satellite attitude control problem.