Artificial Life and Robotics最新文献

Dark colored pigs (Berkshire, Duroc, etc.) are widely recognized nationwide in Japan for their exceptional taste, with the southern Kyushu region being a renowned production area for these esteemed breeds. However, estimating the weight of these pigs using a camera presents a unique challenge. The key process in a camera-based weight estimation system is the precise extraction of the target pig from the background. Typically, cameras capture images from above, as the top-view images provide the most specific growth indicators. However, the image from above contains a ground image. Since Berkshire and Duroc pigs are black and red, respectively, they blend into the ground, making it difficult to accurately segment the pigs in the images. Thus, it is crucial to perfectly distinguish between the ground and the pigs. Therefore, a new extraction method is proposed to distinguish between the ground and pigs by converting depth data based on the pig's position. To enhance the efficiency of pig farming and alleviate the burden on workers, our goal is to develop a system that automatically measures the weight of Berkshire pigs for shipment without background interference. In this study, we installed the system at a Berkshire pig farm and demonstrated the effectiveness of this innovative extraction method for camera-based weight estimation.

Autonomous robots that rely on sensors for operation require fail-soft strategies to continue tasks despite partial sensor failures. We propose a sensor anomaly detection method that monitors changes in sensor data correlations. Our method eliminates the need for pre-defined programming to determine abnormal states for each individual sensor. Furthermore, real-time anomaly detection is possible through sparse structure learning. In the experiment, we evaluated this method on a quadruped robot in a simulated environment. We perturbed the sensor readings by adding two types of large or small noise at one of the robot’s leg joints. When an anomaly was detected, the robot estimates the actual value of the noisy joint using a pre-trained multiple regression model. With our proposed anomaly detection method, the robot successfully completed the walking task in most trials. Specifically, without anomaly detection, adding a large noise to any of the twelve joints resulted in a 0 % success rate. However, with anomaly detection, the success rate improved to over 89 % in seven of the twelve joints.

This study models garbage collection in a local city in Hokkaido, Japan, driven by the increasing burden of collection costs despite a declining population. A unique problem in this city is the large number of garbage stations, which exacerbates the collection burden. We examine the impact of waste volume fluctuations and the number and layout of garbage stations on collection routes and costs to find solutions to this issue. This research aims to develop cost-effective and feasible garbage collection strategies to support evidence-based policymaking. We formulated a garbage collection challenge using mixed integer linear programming to minimize travel distances and operational burdens within vehicle capacity constraints. Numerical simulations reveal significant findings: (i) optimized routes reduce total travel distance by (sim)25% compared to existing routes, (ii) increased waste volumes lead to non-linear increases in route lengths, and (iii) the aggregation strength of garbage stations significantly impacts route efficiency and the number of required stations. Conclusively, this study provides empirical evidence to guide policymakers in optimizing garbage collection systems, ensuring effective resource utilization and maintaining service quality.

In this study, we verify whether computer-generated imagery (CGI) for real image domain adaptation using a generative adversarial network (GAN) based method can improve the accuracy of wood ear mushroom detection. Automated harvesting is important to address labor shortages in agriculture. However, a significant amount of training data is needed to detect crops with features such as wood ear mushroom. In our previous work, a slight improvement in accuracy was achieved with a small number of real-world images by incorporating CGI into the training data. Nevertheless, the reality gap between the CGI and real-world images is expected to degrade detection accuracy. GAN-based methods are effective in bridging the gap, and this study validates this approach. As a result of the experiment, the accuracy improved by approximately 6% in the F2-score between the CGI and GAN-generated datasets. However, there was no significant difference in detection accuracy when training with a combination of real-world and generated image data. In conclusion, this experiment suggests that data augmentation with generated images is not effective for this specific detection target. Nevertheless, this approach should be explored further by generating data on a larger scale with greater computational resources.

Chasing multiple escapees by a group of chasers is an important problem for many living animal species and for various agent systems. On this group chase problem, it has been reported that having two distinct types of chasers in the group, namely diligent and totally lazy chasers, can improve the efficiency of the group of catching all the targets. In this paper, we search for a better group structure by letting each agent have moderate laziness. We find that there exists an optimal group structure, which performs better than the previously reported group which consists of binary (fully diligent and totally lazy) types.

Modular robots, capable of reconfiguration using heterogeneous modules, are considered to be suitable for the completion of base construction tasks on the lunar surface. To this end, there is a need for developing a connector mechanism that connects modules mechanically and electrically in a robust manner. This study developed such a connector with small size, diameter of 13 cm, and satisfactory strength with the ability to withstand a load of 35 Nm. The connector is designed to be genderless, which allows lower restriction in the connection orientation and high flexibility in the reconfiguration of modular robots. The electrical connection is equipped with electrical contact pins and pads that can establish 1000BASE–T Ethernet communication and RS485 communication, and provide 12 V power supply, enabling wired communication between the module computers and the powering of passive modules without a computer. In this study, the developed connectors were mounted onto actual modular robots and their performance and effectiveness was verified. The 16 connectors were sufficiently strong to configure the model of a 40 kg moving robot that consisted of four legs, one body, and four wheels, and the communication quality was acceptable.

Does the enemy of my enemy become my friend? A growing body of literature on structural analysis of interstate relationships has tackled this old question from the complex system and network perspective. However, the mechanism of long-term change in the structure of cooperation and enmity has yet to be fully understood. In search for a general explanation for the long-term evolution of interstate structure, we empirically examine the structural balance theory, which predicts that a signed network evolves toward a more “balanced" structure where in many triangular relationships (i.e., triads) two states tend to share a common enemy or three states cooperate with each other. We investigate the network of alliances (positive edges) and rivalries (negative edges) between sovereign states and examine whether its evolution from 1816 to 2009 can be explained by the structural balance theory. We find the consistency with the structural balance theory drastically changes over time. The empirical pattern follows the prediction by the theory before German unification in the nineteenth century and after World War II while being inconsistent in the middle period. This result reveals the impact of the two historical events on the underlying mechanism of network evolution. Moreover, the contrast with previous studies of signed social networks that generally support the structural balance theory indicates that international alliance and rivalry networks can be promising material to study novel mechanisms behind the time evolution of signed networks and complex systems of conflict and cooperation.

This paper describes a 3D point cloud segmentation pipeline that contributes to the efficiency of decommissioning works at the Fukushima Daiichi Nuclear Power Station. For decommissioning works, simulations and calculations for preliminary work planning using 3D structural models are crucial from a safety and efficiency viewpoint. However, 3D modeling works typically require high costs. Therefore, we aim to improve the efficiency of 3D modeling by segmenting geometric shape regions into categories in a 3D point cloud state using deep learning. Our pipeline uses 3D computer-aided design semantics to create a training dataset that reduces annotation costs and helps learn human knowledge. Performance evaluation results show that the discriminator can discriminate major structural categories with high accuracy using deep learning models. However, we confirm that even the state-of-the-art model has limitations in discriminating structures containing similar shapes between categories and structures in categories with a small number of training data. In the analysis of evaluation results, we discuss challenges encountered by our pipeline for practical applications.

Ear electroencephalogram (ear-EEG) records electrical signals around the ear, offering a more casual and user-friendly approach to EEG measurement. Steady-state visual evoked potential (SSVEP) are brain responses elicited by gazing at flickering stimuli. Ear-EEG can enhance comfort in SSVEP-based brain–computer interface (SSVEP-BCI), but its performance is typically low behind traditional SSVEP-BCI. Additionally, predicting the performance of ear-EEG SSVEP-BCIs before experimentation is challenging, often increasing design costs. This study proposes the SSVEP ratio as a supplementary index to traditional metrics such as information transfer rate (ITR) and BCI accuracy. Using the SSVEP ratio and the KNN algorithm, we predicted BCI accuracy and ITR, aiming to lower design costs. The developed four-inputs ear-EEG SSVEP-BCI achieved a maximum BCI accuracy of 89.17 ± 3.62% and an ITR of 10.60 ± 0.36 bits/min. Predicted BCI accuracy was 90.21 ± 3.25% and an ITR was 9.43 ± 0.96 bits/min in ear-EEG SSVEP-BCI. Predicted values matched the actual results, demonstrating that the SSVEP ratio can effectively predict BCI accuracy, thereby streamlining the design process for ear-EEG SSVEP-BCI.

Recently, many researchers have expected millimeter-sized microrobots to work in narrow spaces. However, it is challenging to integrate the actuators, controllers, sensors, and energy sources into millimeter-sized microrobots. A small actuator with low power consumption is required to realize millimeter-sized microrobots. Previously, the authors developed a new linear electrostatic motor for microrobots. However, most microrobots rely on rotary actuators to expand their application scenarios and enhance adaptability. In this paper, the authors designed and developed a rotary-type electrostatic motor to provide a low-power drive solution for microrobots to address the limitations of linear motors and broaden their range of applications. Through experimentation, we identified an issue with reverse rotation in the electrostatic motor and analyzed its causes. To address the reverse-rotation issue, we proposed improvements, including optimizing the electrode structure and adjusting the drive waveform, which significantly enhanced the stability of forward rotation. The author plans to refine the motor's design further and integrate it into a microrobot system.