Artificial Life and Robotics最新文献

In this paper, we propose a three-layered neural network controller (NC) optimized using an improved bat algorithm (BA) for a rotary crane system. In our previous study, the simulation results showed that an NC optimized using the original BA exhibits good control and evolutionary performance. However, the simulation execution time was long. Therefore, to address this problem, we propose an improved BA that reduces the execution time. We show that the NC optimized by the improved BA exhibits the same control performance as that optimized via conventional methods. It is also shown that the time for evolutionary calculations can be reduced.

We study the data of Japanese video-sharing platform, Niconico, which contains 21 million videos. From our analysis, the rank size distribution of video views is found to exhibit a crossover from a power law with an exponent around (-0.5) for the top (approx 10^5) movies to another power low with exponent around (-1) for the movies in the following ranks. The probability density function of video views for the bottom (90%) movies is well fitted by log-normal distribution. This implies that, while videos in the top rank regime follow a different dynamics which yields the power law, videos in the middle and low rank regime seem to be evolving according to a random multiplicative process. Furthermore, we observe temporal relaxation process of video views for 3 years. Temporal relaxation process of video views is grouped by the size of the number of video views, and averaged within each size group. Interestingly, the daily video views universally show power-law relaxation in all view size, from the top total view group ((10^6-10^7)) to the bottom group ((approx 10^2)). This indicates the existence of memory processes longer than the exponential function, which are universally independent of video size.

We study coupling complexity in multivariate time series generated by echo state networks subject to i.i.d. input signals using the ordinal persistent index as a coupling complexity measure. Coupling complexity is a notion of complexity focusing on the relations among components of a given system. Given a time segment of a multivariate time series, its ordinal persistent index is defined by taking the persistent homology of a filtered simplicial complex reflecting similarity among the ordinal patterns of individual time series. As the strength of input signals increases, the dynamics of echo state networks shift from asynchronous ones to more synchronized ones. We show that the original ordinal persistent index cannot capture such change in the synchronization behavior, but a generalized version of the ordinal persistent index is sensitive to the change: the latter takes relatively high values between the two extremes, namely when the strength of input signals to the echo state networks is within a certain range of intermediate values.

Advances in neural network (NN) models and learning methods have resulted in breakthroughs in various fields. A larger NN model is more difficult to install on a computer with limited computing resources. One method for compressing NN models is to quantize the weights, in which the connection weights of the NNs are approximated with low-bit precision. The existing quantization methods for NN models can be categorized into two approaches: quantization-aware training (QAT) and post-training quantization (PTQ). In this study, we focused on the performance degradation of NN models using PTQ. This paper proposes a method for visually evaluating the performance of quantized NNs using topological data analysis (TDA). Subjecting the structure of NNs to TDA allows the performance of quantized NNs to be assessed without experiments or simulations. We developed a TDA-based evaluation method for NNs with low-bit weights by referring to previous research on a TDA-based evaluation method for NNs with high-bit weights. We also tested the TDA-based method using the MNIST dataset. Finally, we compared the performance of the quantized NNs generated by static and dynamic quantization through a visual demonstration.

The translation activity of language is a link and bridge for the integration of politics, economy, and culture in various countries. However, manual translation requires high quality of professional translators and takes a long time. The study attempts to introduce dual learning on the basis of traditional neural machine translation models. The improved neural machine translation model includes decoding of the source language and target language. With the help of the source language encoder, forward translation, backward backtranslation, and parallel decoding can be achieved; At the same time, adversarial training is carried out using a corpus containing noise to enhance the robustness of the model, enriching the technical and theoretical knowledge of existing neural machine translation models. The test results show that compared with the training speed of the baseline model, the training speed of the constructed model is 115 K words/s and the decoding speed is 2647 K words/s, which is 7.65 times faster than the decoding speed, and the translation quality loss is within the acceptable range. The mean bilingual evaluation score for the “two-step” training method was 16.51, an increase of 3.64 points from the lowest score, and the K-nearest-neighbor algorithm and the changing-character attack ensured the semantic integrity of noisy source language utterances to a greater extent. The translation quality of the changing character method outperformed that of the unrestricted noise attack method, with the highest bilingual evaluation study score value improving by 3.34 points and improving the robustness of the model. The translation model constructed by the study has been improved in terms of training speed and robustness performance, and is of practical use in many translation domains.

Visual simultaneous localization and mapping (SLAM) is a critical technology for robots to perform high-precision navigation, increasing the focus among researchers to improve its accuracy. However, improvements in SLAM accuracy always come at the cost of an increased memory footprint, which limits the long-term operation of devices that operate under constrained hardware resources. Application of quantization methods is proposed as a promising solution to this problem. Since quantization can result in performance degradation, it is crucial to quantitatively evaluate the trade-off between potential degradation and memory savings to assess its practicality for visual SLAM. This paper introduces a mechanism to evaluate the influence of a quantization method on visual SLAM, and applies it to assess the impact of three different quantization methods on ORB-SLAM3. Specifically, we examine two static quantization methods and a dynamic quantization method called error diffusion, which can pseudo-preserve image shading information. The paper contributes to the conclusion that error diffusion, with controlled weight parameters in the error diffusion filter, can suppress degradation and reduce the memory footprint, demonstrating its effectiveness in dynamic environments.

The integration of modern manufacturing systems has promised increased flexibility, productivity, and efficiency. In such an environment, collaboration between humans and robots in a shared workspace is essential to effectively accomplish shared tasks. Strong communication among partners is essential for collaborative efficiency. This research investigates an approach to non-verbal communication cues. The system focuses on integrating human motion detection with vision sensors. This method addresses the bias human action detection in frames and enhances the accuracy of perception as information about human activities to the robot. By interpreting spatial and temporal data, the system detects human movements through sequences of human activity frames while working together. The training and validation results confirm that the approach achieves an accuracy of 91%. The sequential testing performance showed an average detection of 83%. This research not only emphasizes the importance of advanced communication in human–robot collaboration, but also effectively promotes future developments in collaborative robotics.

In this study, facial skin temperature distribution (FSTD) is focused on as a new driver monitoring index. FSTD is an autonomic index that can be measured remotely. Studies have been conducted to estimate drowsiness based on FSTD using modelng methods such as CNN, a type of deep learning, and sparse modeling, which can be trained with a small amount of data. These studies, however, only evaluated front-facing facial thermal images. FaceMesh is a model that extracts 478 3D facial feature landmarks from a 2D face image. In contrast to conventional models that extract only 68 facial feature landmarks, FaceMesh can extract facial feature landmarks for the entire face, including the cheeks, forehead, and other areas of the face that are in the blind spots. This study aims to improve the accuracy of drowsiness estimation by applying FaceMesh to automatically detect tilted faces and not including tilted images in the training data. As a result, the method proposed in this study improved drowsiness estimation accuracy by about 6% compared to the old method, which did not take face orientation into account.

There is a growing expectation that deep reinforcement learning will enable multi-degree-of-freedom robots to acquire policies suitable for real-world applications. However, a robot system with a variety of components requires many learning trials for each different combination of robot modules. In this study, we propose a hierarchical policy design to segment tasks according to different robot components. The tasks of the multi-module robot are performed by skill sets trained on a component-by-component basis. In our learning approach, each module learns reusable skills, which are then integrated to control the whole robotic system. By adopting component-based learning and reusing previously acquired policies, we transform the action space from continuous to discrete. This transformation reduces the complexity of exploration across the entire robotic system. We validated our proposed method by applying it to a valve rotation task using a combination of a robotic arm and a robotic gripper. Evaluation based on physical simulations showed that hierarchical policy construction improved sample efficiency, achieving performance comparable to the baseline with 46.3% fewer samples.