arXiv subjects

Toshiyuki Oshima

Publications and source records attributed to Toshiyuki Oshima.

4 recordsLinked to original sources

arXiv2025

Constraint-Driven Multi-USV Coverage Path Generation for Aquatic Environmental Monitoring

In this article, we address aquatic environmental monitoring using a fleet of unmanned surface vehicles (USVs). Specifically, we develop an online path generator that provides either circular or elliptic paths based on the real-time feedback so that the USVs efficiently sample the sensor data over given aquatic environment. To this end, we begin by formulating a novel online path generation problem for a group of Dubins vehicles in the form of cost minimization based on the formulation of persistent coverage control. We then transform the cost minimization into a constraint-based specification so that a prescribed performance level is certified. An online coverage path generator is then designed based on the so-called constraint-based control in order to meet the performance certificate together with additional constraints inherent in the parameters that specify the paths. It is also shown there that the present constraint-based approach allows one to drastically reduce the computational complexity stemming from combinations of binary variables corresponding to the turning directions of the USVs. The present coverage path generator is finally demonstrated through simulations and experiments on an original testbed of multiple USVs.

eess.SY↗

arXiv2024

Distributed Shape Learning of Complex Objects Using Gaussian Kernel

This paper addresses distributed learning of a complex object for multiple networked robots based on distributed optimization and kernel-based support vector machine. In order to overcome a fundamental limitation of polynomial kernels assumed in our antecessor, we employ Gaussian kernel as a kernel function for classification. The Gaussian kernel prohibits the robots to share the function through a finite number of equality constraints due to its infinite dimensionality of the function space. We thus reformulate the optimization problem assuming that the target function space is identified with the space spanned by the bases associated with not the data but a finite number of grid points. The above relaxation is shown to allow the robots to share the function by a finite number of equality constraints. We finally demonstrate the present approach through numerical simulations.

cs.RO↗

arXiv2024

Clustering-friendly Representation Learning for Enhancing Salient Features

Recently, representation learning with contrastive learning algorithms has been successfully applied to challenging unlabeled datasets. However, these methods are unable to distinguish important features from unimportant ones under simply unsupervised settings, and definitions of importance vary according to the type of downstream task or analysis goal, such as the identification of objects or backgrounds. In this paper, we focus on unsupervised image clustering as the downstream task and propose a representation learning method that enhances features critical to the clustering task. We extend a clustering-friendly contrastive learning method and incorporate a contrastive analysis approach, which utilizes a reference dataset to separate important features from unimportant ones, into the design of loss functions. Conducting an experimental evaluation of image clustering for three datasets with characteristic backgrounds, we show that for all datasets, our method achieves higher clustering scores compared with conventional contrastive analysis and deep clustering methods.

cs.CV↗

arXiv2013

An Effective Potential Theory for Transport Coefficients Across Coupling Regimes

A plasma transport theory that spans weak to strong coupling is developed from a binary collision picture, but where the interaction potential is taken to be an effective potential that includes correlation effects and screening self-consistently. This physically motivated approach provides a practical model for evaluating transport coefficients across coupling regimes. The theory is shown to compare well with classical molecular dynamics simulations of temperature relaxation in electron-ion plasmas, as well as simulations and experiments of self-diffusion in one component plasmas. The approach is versatile and can be applied to other transport coefficients as well.

physics.plasm-ph↗