Artificial Life and Robotics最新文献

Unlike the conventional swarm or evolutionary optimizations that are generally assumed the “pre-defined” bounded search space, this paper addresses the optimization for the “unbounded” search space. For this purpose, this paper proposes novelty-based multi-objectivization with local and rough area search (NM-LRS), which adds the novelty criterion in the given optimization criteria to roughly search the unbounded search space for obtaining the “potential area” where the optimal solution is most likely located and then searches the “potential area” to find the optimal solution by a local area search. To investigate the effectiveness of the proposed methods, the experiment compares the proposed methods with the conventional optimization methods for the unbounded multi-modal optimization and has revealed the following implications: (i) the peak ratio (i.e., the ratio of the founded peaks of the multi-modal function) of NM-LRS is higher than that of the conventional methods; and (ii) NM-LRS is robust for the location of the initial search area in the most functions.

Underwater environments provide significant challenges for humans, thus researchers have focused on controlling underwater robots equipped with manipulators known as Underwater Vehicle-Manipulator System (UVMS) that perform underwater tasks instead of humans. To achieve high-precision control of UVMS, an accurate mathematical model must be developed. However, there are modeling errors between the UVMS model used for control system and the fluid forces that actually act on the robot. In conventional studies, control methods based on joint space have been used as a compensation controller for disturbances, including modeling errors. This paper proposes a Resolved Acceleration Control (RAC) method for UVMS that incorporates a Model Error Compensator (MEC), a control method based on task space, designed to minimize these model errors. The proposed method aims to achieve robust trajectory tracking control for UVMS by suppressing the uncertainties in modeling of fluid forces and the effects of disturbances. Furthermore, unlike many prior studies that demonstrate the effectiveness of their methods through simulations, this study validates the proposed method through position control experiments of a robot under wave disturbances. The experimental results confirm the robustness of the control system against modeling errors and wave disturbances, demonstrating the usefulness of the proposed method.

People with serious physical disabilities (ex. spinal muscular atrophy) find it difficult to control a robot wheelchair. Although gesture-based robot control mechanisms have been proposed, making such gestures is not always feasible. To this end, this paper proposes a brain–machine interface (BMI) for robot control by processing electroencephalograph (EEG) signals captured from non-invasive external device. We systematically process the EEG signals to first estimate the most prominent brain channels. This eliminates the redundant information or noise which adversely influences the recognition accuracy. We then estimate the most prominent EEG waves among the prominent channels. Later, the combination of prominent brain waves among the prominent channels which gives the most accurate robot control are estimated. Convolutional neural network (CNN) is used to process the EEG signals. The user can control the robot in four different directions. Experiments with actual external BMI device are performed and robot is controlled.

This paper proposes an improvement of SimCLR for defect recognition tasks by image synthesis using weighted averages. There are studies on applying contrastive learning to defect detection in industrial products. This is because the number of defective products is quite small compared to non-defective products, and contrastive learning is a method that allows you to train a model with a small dataset by augmenting images and comparing them. However, problems with random trimming have been reported for the combination of defect detection and contrastive learning. Since defect images consist of defect areas and non-defect areas, augmentation by random cropping does not work well. To solve this problem, this study proposes the addition of image synthesis using weighted averaging to the conventional SimCLR’s augmentation method. The proposed method avoids wasteful learning that attracts feature vectors between cropped defect and non-defect areas. In the experiment, a CNN was trained on a small dataset of 32 images, and our proposed method improved AUC by 15% compared to the conventional method.

In this study, we propose a method for learning a latent space representing 6-DoF poses and performing 6-DoF control in the latent space using NewtonianVAE. NewtonianVAE, a type of world models based on Variational Autoencoder (VAE), can learn the dynamics of the environment as a latent space from observational data and perform proportional control based on the estimated position on the latent space. However, previous research has not demonstrated 6-DoF pose estimation and control using NewtonianVAE. Therefore, we propose 6D NewtonianVAE, which extends the latent space by incorporating the rotation vector to construct the latent space representing 6-DoF poses and perform 6-DoF control based on the estimated poses. Experimental results showed that our method achieves 6-DoF control with an accuracy within 7 mm and 0.02 rad in a real-world. It was also shown that 6-DoF control is possible even in unseen environments. Our approach enables end-to-end 6-DoF pose estimation and control without annotated data. It also eliminates the need for RGB-D or point cloud data and relies solely on RGB images, reducing implementation and computational costs.

Black box variational inference is a promising framework in a succession of recent efforts to make Variational Inference more “black box”. However, in its basic version it either fails to converge due to instability or requires some fine-tuning of the update steps prior to execution that hinders it from being completely general purpose. We propose a method for regulating its parameter updates by re-framing stochastic optimization as a multivariate estimation problem. Borrowing from estimation theory, we examine the properties of the James–Stein estimator as a replacement for the arithmetic mean of Monte Carlo estimates of the gradient of the evidence lower bound. Theoretical guarantees for its variance reduction properties are also given. We show through simulations that the proposed method provides relatively weaker variance reduction than Rao-Blackwellization, but offers a tradeoff of being simpler and requiring no prior analysis on the part of the user. Comparisons on benchmark datasets also demonstrate a consistent performance at par or better than the Rao-Blackwellized approach in terms of resulting model fit.

Here, we concentrate on the world that only chemicals are allowed to use as cues from agents, the chemicals secreted from all agents, diffuse and decay under fluid conditions, give rise to change of motility to agents, that is called chemotaxis. At first, motility of single agent is confirmed, and then we show a simple mechanism of predator (chaser)–prey (target) system consist of such chemotactic agents only. Finally, we explicitly consider fluid conditions in the system. The model system has parameter (alpha), corresponding diffusion coefficient of the chemicals, inversely relates to Péclet numbers. The smaller Péclet numbers give rise to more obscure chemical traces, but leading to higher survivability-efficient to predator (chaser) as well as prey (target), indicating that they can use complex traces to change their moving directions without using any waves, such as electromagnetic and/or sound. These results can be regarded as an emergence phenomena of diffusion- and chemotaxis-driven swarm intelligence.

This paper proposes a new multi-objective reinforcement learning (MORL) algorithm for robotics by extending policy improvement with path integral ((text {PI}^2)) algorithm. For a robot motion acquisition problem, most existing MORL algorithms are hard to apply, because of the high-dimensional and continuous state and action spaces. However, policy-based algorithms such as (text {PI}^2) can be applied to solve this problem in single-objective cases. Based on the similarity of (text {PI}^2) and evolution strategies (ESs) and the fact that ESs are well-suited for multi-objective optimization, we propose an extension of (text {PI}^2) and some techniques to speed up the learning. The effectiveness is shown via numerical simulations.

Steady-state visual evoked potential (SSVEP) brain-computer interface (BCI) are known for high speed, accuracy, and multivalue input. Integrating ear-electroencephalogram (EEG) can make SSVEP-BCI more accessible for everyday use. This study introduces a reliability score to enhance the performance of ear-EEG SSVEP-BCI by dynamically adjusting measurement duration and enabling asynchronous detection. Two analysis methods, learning canonical correlation analysis (LCCA) and task-related component analysis, were evaluated. Using the reliability score, the accuracy for ear-EEG SSVEP-BCI reached (100)% with an information transfer rate (ITR) of (22.36pm 3.54) bits/min, compared to (61.93pm 9.22)% accuracy and (15.32pm 4.59) bits/min ITR without the reliability score. These findings demonstrate that the reliability score significantly improves ear-EEG SSVEP-BCI performance, suggesting its potential to enhance usability in practical applications.

This paper focuses on the collision avoidance of multiple UAVs using collision cones (CCs) and control barrier functions (CBFs). Each UAV is separately controlled toward a given goal while avoiding collision with other UAVs, which are considered moving obstacles. We first propose a new collision avoidance control method based on CCs and CBFs without numerical optimization. This method significantly lowers computational costs compared to existing optimization-based approaches. In addition, we propose a new optimization-based method using CCs and CBFs. A key feature of the proposed method is that the desired control input used in numerical optimization is modified based on CCs and CBFs, in contrast to existing methods that use a desired control input designed without considering obstacles. We evaluate and compare the effectiveness of the proposed methods through extensive simulations. Experimental results using real quadrotors are also shown.