This paper proposes a novel algorithm, Terrain Analysis–Rapidly exploring Random Tree* (TA-RRT*), which improves the performance of path planning using adaptive sampling and step size based on terrain analysis. While existing RRT*-based algorithms perform node sampling and tree expansion in various ways to optimize path planning, they may still generate inefficient paths in complex terrain environments. In this paper, TA-RRT* analyzes the complexity of the terrain to generate a probability density function, which is then used to guide adaptive sampling and tree expansion during the path planning process. In addition, we reduce the length of generated paths using an adaptive step size based on terrain complexity. To verify the performance of the proposed algorithm, we conduct experiments with various path planning algorithms in three different environments: simple, intermediate, and complex. The experimental results demonstrate that TA-RRT* outperforms other algorithms in terms of path length, computational time, and memory usage. Furthermore, its robustness was validated in dynamic environments, where it effectively performed real-time path replanning to adapt to environmental changes, such as the appearance of new obstacles.

본 논문에서는 비행 중 비행체 내·외부에서 발생하는 동적상황 변화에 대해 대응 우선순위에 따라 무인기가 자율적으로 대처할 수 있는 무인기 자율 임무관리 소프트웨어의 설계 및 검증 기법을 제안한다. 제안된 무인기 자율 임무관리 소프트웨어는 다중 스레드 기반의 동시 수행 가능한 모듈화 구조로 하위 소프트웨어들을 설계하였으며, 이를 검증하기 위해 1) 모든 동적상황 변화에 대한 스레드들 간의 요청/응답 메시지를 확인하는 소프트웨어 통합 기법과 2) 소프트웨어 통합검증환경을 활용하여 소프트웨어의 기능 검증을 수행하는 통합시험 및 3) 정량적 성능 검증을 수행하는 성능시험 기법을 제시한다. 특히, 통합시험과 성능시험을 수행하기 위해 소프트웨어 통합검증환경을 제작하여 활용한다.

Histopathological whole slide images (WSIs) provide critical information for disease diagnosis, yet their interpretation remains a time-consuming and expertise-dependent process. Recent advances in deep learning have shown promise in automating and improving histopathological analysis; however, the performance of convolutional neural networks (CNNs) and Vision Transformers (ViTs) across multi-organ disease classification remains insufficiently explored. In this study, we applied CNN- and ViT-based classification models to WSIs of major human organs, including the heart, lung, liver, and pancreas. Both single-organ models (classifying normal versus diseased tissue per organ) and integrated multi-organ models (classifying across organs within a unified framework) were evaluated. Model performance was compared in terms of classification accuracy. Among the ViT-based models, Swin-T achieved the highest accuracy (0.9964 ± 0.0036), while DenseNet-161 outperformed other CNN-based models (0.9811 ± 0.0078). Importantly, no substantial drop in classification performance was observed when extending from single-organ to multi-organ disease classification. These findings demonstrate that deep learning classification models based on ViTs and CNNs, particularly Swin-T and DenseNet-161, achieve robust performance in multi-organ histopathological disease classification. The results highlight their potential utility for the development of generalized diagnostic support software for multi-organ disease classification in histopathology and their future integration into routine digital pathology practice.

This paper presents XGate-Net, a lightweight multi-task perception framework designed to perform semantic segmentation and monocular depth estimation simultaneously for autonomous driving. Conventional multi-task structures that rely on fully shared backbones or simple feature fusion often suffer from negative transfer, which may degrade task-specific feature representations. To address these limitations, XGate-Net introduces three core components. First, a bidirectional cross-task gating fusion framework selectively regulates information exchanged between the two task branches, enabling controlled refinement while preserving task-specific characteristics. Second, a task-specific dynamic pyramid fusion module adaptively aggregates multi-scale features through scale-aware attention, enhancing boundary precision in segmentation and geometric consistency in depth estimation. Third, an edge–region cross feedback mechanism leverages complementary edge cues from depth and region cues from segmentation to improve inter-task consistency and prediction stability. Experiments on the VKITTI2–Sunset dataset demonstrate that XGate-Net delivers segmentation and depth estimation performance competitive with single-task baselines, while maintaining real-time inference and low computational cost. Additional evaluations on the NVIDIA Jetson AGX Orin platform further confirm the model’s suitability for embedded deployment. These results indicate that structured and selective cross-task interaction offers an effective approach for efficient multi-task perception in autonomous driving applications.