The provision of audio augmented reality (AAR) experiences is becoming more widespread.In this study, to investigate the influence of device design on AAR experience from the perspective of acoustic transparency, physical and subjective evaluations were conducted using five devices with different shapes and transparency modes.In the subjective evaluation, perceived transparency, impressions of real-world sound, and subjective impressions of AAR experience when wearing each device were evaluated for two distinct content types.We found that device design can potentially influence impressions of real-world sound, such as auditory source width, listener envelopment and punch, and subjective impressions during AAR experience.Devices with high transparency were more likely to draw attention to real-world sounds when users were experiencing AAR, and the experience was evaluated as enjoyable and natural.Two demonstration experiments showed that adding virtual sounds by open-ear earphones to real contents can provide acoustic effects such as distance enhancement.

With the continuous shrinking of the semiconductor device dimensions, mask topography effects stand out among the major factors influencing the lithography process. Including these effects in the lithography optimization procedure has become necessary for advanced technology nodes. However, conventional rigorous simulation for mask topography effects is extremely computationally expensive for high accuracy. In this work, we propose TEMPO as a novel generative learning-based framework for efficient and accurate 3D aerial image prediction. At its core, TEMPO comprises a generative adversarial network capable of predicting aerial image intensity at different resist heights. Compared to the default approach of building a unique model for each desired height, TEMPO takes as one of its inputs the desired height to produce the corresponding aerial image. In this way, the global model in TEMPO can capture the shared behavior among different heights, thus, resulting in smaller model size. Besides, across-height information sharing results in better model accuracy and generalization capability. Our experimental results demonstrate that TEMPO can obtain up to 1170x speedup compared with rigorous simulation while achieving satisfactory accuracy.

Sufficient supervised information is crucial for any machine learning models to boost performance. However, labeling data is expensive and sometimes difficult to obtain. Active learning is an approach to acquire annotations for data from a human oracle by selecting informative samples with a high probability to enhance performance. In recent emerging studies, a generative adversarial network (GAN) has been integrated with active learning to generate good candidates to be presented to the oracle. In this paper, we propose a novel model that is able to obtain labels for data in a cheaper manner without the need to query an oracle. In the model, a novel reward for each sample is devised to measure the degree of uncertainty, which is obtained from a classifier trained with existing labeled data. This reward is used to guide a conditional GAN to generate informative samples with a higher probability for a certain label. With extensive evaluations, we have confirmed the effectiveness of the model, showing that the generated samples are capable of improving the classification performance in popular image classification tasks.

Lithography simulation is one of the key steps in physical verification, enabled by the substantial optical and resist models. A resist model bridges the aerial image simulation to printed patterns. While the effectiveness of learning-based solutions for resist modeling has been demonstrated, they are considerably data-demanding. Meanwhile, a set of manufactured data for a specific lithography configuration is only valid for the training of one single model, indicating low data efficiency. Due to the complexity of the manufacturing process, obtaining enough data for acceptable accuracy becomes very expensive in terms of both time and cost, especially during the evolution of technology generations when the design space is intensively explored. In this paper, we propose a new resist modeling framework for contact layers, utilizing existing data from old technology nodes and active selection of data in a target technology node, to reduce the amount of data required from the target lithography configuration. Our framework based on transfer learning and active learning techniques is effective within a competitive range of accuracy, i.e., $3 \times -10 \times $ reduction on the amount of training data with comparable accuracy to the state-of-the-art learning approach.

Lithography simulation is one of the key steps in physical verification, enabled by the substantial optical and resist models. A resist model bridges the aerial image simulation to printed patterns. While the effectiveness of learning-based solutions for resist modeling has been demonstrated, they are considerably data-demanding. Meanwhile, a set of manufactured data for a specific lithography configuration is only valid for the training of one single model, indicating low data efficiency. Due to the complexity of the manufacturing process, obtaining enough data for acceptable accuracy becomes very expensive in terms of both time and cost, especially during the evolution of technology generations when the design space is intensively explored. In this work, we propose a new resist modeling framework for contact layers that utilizes existing data from old technology nodes to reduce the amount of data required from a target lithography configuration. Our framework based on residual neural networks and transfer learning techniques is effective within a competitive range of accuracy, i.e., 2-10X reduction on the amount of training data with comparable accuracy to the state-of-the-art learning approach.

A proof-of-concept non-binary cyclic ADC with proposed correlated level shifting (CLS) technique is designed and fabricated in 90nm CMOS technology. By applying the odd/even structure to a multiplying digital-to-analog converter (MDAC), the amplifier with CLS can be used for a cyclic ADC, so that the allowable dynamic range of the proposed cyclic ADC almost is doubled comparing to the conventional cyclic ADC. As a result, the SNDR of ADC can be improved with smaller power penalty. Measurement results of the prototype confirm the effectiveness of the proposed CLS technique for cyclic ADC.

This paper presents a prototype of 14bit 80kSPS non-binary cyclic ADC without high accuracy analog components and complicated digital calibration. Since the redantancy of non-binary ADC tolerates the non-idealities such as capacitor mismatch and finite amplifier gain, the design consideration of this high accuracy ADC are be focused on the capacitance of sampling capacitor to satisfy the overall kT/C noise target, the drivability and linearity of amplifier. The proposed proof-of-concept cyclic ADC is designed and fabricated in TSMC 90nm CMOS technology. Peak SNDR×81.9dB is achieved while Fs×80kSPS with a poor gain of the amplifier as low as 66dB dissipating 8mW at VDD×3.3V in analog circuits.

Autonomous decentralized control is a key concept for the realization of highly adaptive behavior. However, universal design of autonomous decentralized control that ensures rich adaptability is still lacking. In this study, we tackle this problem through the development of a two-dimensional sheet-like robot, SheetBot. The SheetBot is a suitable model system for the establishment of universal design principles for autonomous decentralized control, because it can bend reasonably to the circumstances encountered due to its large surface area, and also because its applications are widely expected. To realize highly adaptive locomotion with SheetBot, we are inspired by the decentralized control scheme of the scaffold-based locomotion of snakes. We extend this design scheme to a two-dimensional bodily structure on the basis of a continuum model. Simulation results show that SheetBot can locomote on various kinds of irregular terrain with minimal control inputs by implementing the proposed autonomous decentralized control scheme.

Autonomous decentralized control is an attractive concept for designing robots that can exhibit highly adaptive animal-like behaviour. Despite its appeal, previous robots could only adapt to a limited number of environments. To tackle this problem, we hypothesize that robots will move in various environments by simply implementing a “scaffold-exploitation mechanism”. In our previous work, we drew our inspiration from flatworms, which move over various terrains using their two-dimensional sheet-like body, and designed an autonomous decentralized control scheme for a similar robot, SheetBot. In this study, we extend our previous control scheme so that it can effectively exploit the features in its surrounding environment as scaffolds. The validity of the extended control scheme is confirmed by simulations. Our control scheme is not specific to a certain environment but could be applicable in any environment; it thus forms the basis for developing a multi-terrestrial robot whose working area covers land, sea, and even air.

The development of a robot motion control scheme and a mechanical load adjuster with a motion measurement interface is addressed in this paper. To improve the efficiency of a task involving human-robot cooperation, we designed a novel robot control system, in which a multiple-load state can be provided for a human operator. This system also can provide a multiple-dynamics state during a task involving human-robot dynamical cooperation. The multiple-load state including its transition is effective and efficient in such a task. A single load state can be easily provided by the impedance control of the robot motion thus far; however, the multiple-load state and its transition are difficult to realize using a conventional control scheme. The proposed control scheme differs widely from a conventional impedance control scheme in that the multiple- load state as well as both active and passive states cannot be induced in the single robotic system. Under the proposed control scheme, the load state can be adjusted with the various dynamics in the active or passive state. To confirm the effectiveness of the proposed control system, human-robot cooperative experiments were carried out. Results showed that the proposed control scheme can provide the multiple-load state for use in a human-robot cooperative task system.

This paper presents an arithmetic module generator based on an arithmetic description language called ARITH. The use of ARITH makes it possible to describe a wide variety of arithmetic algorithms in a unified manner. The ARITH descriptions are formally verified in the generator even if the arithmetic algorithms include unconventional number systems for operands or internal variables. The proposed generator also optimizes arithmetic algorithms by using performance profiles derived from the previous generation. From these features, we can obtain high-performance arithmetic modules whose functions are completely verified at the algorithm level. In this paper, we demonstrate that the optimal prefix adders improved the performance of generated arithmetic modules such as multipliers in comparison with the standard prefix adders.

This paper presents a formal approach to verify arithmetic circuits using symbolic computer algebra. Our method describes arithmetic circuits directly with high-level mathematical objects based on weighted number systems and arithmetic formulae. Such circuit description can be effectively verified by polynomial reduction techniques using Grobner Bases. In this paper, we describe how the symbolic computer algebra can be used to describe and verify arithmetic circuits. The advantageous effects of the proposed approach are demonstrated through experimental verification of some arithmetic circuits such as multiply-accumulator and FIR filter. The result shows that the proposed approach has a definite possibility of verifying practical arithmetic circuits where the conventional techniques failed.