The tellability of a car’s style—its capacity to communicate emotions, brand identity, country of origin, and car segment—remains underexplored in user-centered automotive design. This study aims to analyze the tellability of car styles by examining relationships between cars and these factors. A computational framework was employed to collect data from users and a selected set of cars and to quantitatively map perceptual relationships using block-seriation analysis. Three robust perceptual patterns—Stable, Friendly, and Powerful—and their associated Core Emotional Themes were identified: Calm, Grounded Dependability (Stable); Warm, Approachable Comfort (Friendly); and High-Energy Performance and Powerful Presence (Powerful). The research further highlights that brand recognition depends not only on distinctiveness but also on the consistency of design cues, with recognition emerging from visually coherent and differentiated brand elements. Moreover, shared design trends and platform similarities can blur brand boundaries, whereas minor similarities generate perceptual uncertainty across brands. At the country-of-origin level, perceptions exist on a continuum from clear to ambiguous, also reflecting globalization’s influence on stylistic identities. Car segments are often perceived as fuzzy rather than discrete, and recognition strength does not always align with segment clarity. By capturing the complexity of observer perception, these findings advance understanding of how users interpret and evaluate automotive designs, providing actionable insights for design, brand management, and strategic development.
This paper presents NavWareSet , a novel dataset crafted to advance socially compliant robot navigation research. NavWareSet provides multi-modal recordings of both socially compliant and non-compliant robot trajectories in controlled indoor environments. Drawing upon seven carefully selected scenarios, it captures complex human-robot interactions and a range of navigation challenges that mirror realistic social contexts. NavWareSet establishes a rich dataset for evaluating and training navigation algorithms by incorporating two distinct robot platforms—Toyota Human Support Robot (HSR) and Clearpath’s Jackal—and systematically varying their navigation behaviors. With data modalities spanning lidar, RGB-D camera, odometry, and human position annotations, NavWareSet enables fine-grained analysis of the robot’s decision-making process and its impact on human comfort and safety. Ultimately, this dataset provides a versatile resource for developing robust, ethically guided navigation policies and for measuring their performance across a range of social situations. More information can be seen at: https://anr-navware.github.io/navwareset/ .
Hydrogen-based transit networks are emerging as a pillar of decarbonizing the transportation sector, requiring advanced control strategies to ensure dispatchable operation, energy efficiency, and economic viability. To this end, this paper proposes a sophisticated stochastic model-predictive control orchestrator for the coordinated operation of a hydrogen refueling station (HRS) powered by renewable energy sources and on-site hydrogen. The controller optimizes refueling schedules, electrolyzer and compressor operations, and participation in coupled electricity-hydrogen markets while explicitly accounting for stochastic variations in RES generation, energy prices, and hydrogen supply-chain delays. As a key novelty, a probabilistic hydrogen procurement model is introduced to represent delivery-time uncertainty within the predictive horizon. To handle the resulting complexity, a convex relaxation combined with scenario-based decomposition is employed, yielding a formulation that guarantees computational efficiency, recursive feasibility, and near-optimal performance under uncertainty. Simulation studies over multi-day operation horizons show that the proposed strategy ensures more than 98% refueling availability, reduces energy costs by about 15%, and increases total profit by up to 57% while maintaining a 60% reduction in computation time relative to MPC. The integrated system is generic for coordinated control of HRS infrastructures in next-generation sustainable energy systems.
Introduction. The growing need for efficient and high-performance electric drive systems has led to increased research in advanced control strategies for multi-machine configurations. Among them, dual-star permanent magnet synchronous machines (DSPMSMs) connected in parallel to a single inverter offer a promising solution for applications requiring high reliability and precise control. Problem. Conventional direct torque control (DTC) strategies, typically relying on PI controllers, suffer from significant torque and flux ripples, which negatively impact system efficiency and dynamic response. Moreover, these traditional controllers face challenges in handling parameter variations and external disturbances, limiting their applicability in demanding environments. Goal. This study aims to enhance the performance of DSPMSM drive systems by improving speed regulation, minimizing torque and flux fluctuations, and increasing robustness against disturbances, thereby ensuring greater efficiency and stability. Methodology. To address these challenges, we propose a novel DTC strategy that replaces the conventional PI controller with a type-2 fuzzy logic controller (T2-FLC). This intelligent control approach leverages the adaptability of fuzzy logic to improve response accuracy and dynamic performance. The proposed methodology is validated through extensive simulations using MATLAB/Simulink, analyzing various operating conditions and comparing the performance with conventional DTC techniques. Results. Simulation results confirm that the T2-FLC-based DTC significantly reduces torque and flux ripples while ensuring precise speed regulation. The proposed approach also demonstrates improved robustness against disturbances and parameter variations, outperforming traditional PI-based DTC in terms of efficiency and control accuracy. Scientific novelty. This research introduces an innovative application of T2-FLC in DTC for parallel-connected DSPMSMs, offering a novel control strategy that effectively mitigates the drawbacks of conventional methods. The integration of T2-FLC into the DTC framework provides enhanced adaptability and superior performance, distinguishing this study from existing works. Practical value. The proposed control strategy enhances the reliability, efficiency, and stability of DSPMSM-based drive systems, making it well-suited for high-performance applications such as railway traction, electric vehicles, and industrial automation. By improving control precision and robustness, this approach contributes to the advancement of intelligent drive technologies in modern electric propulsion systems. References 39, tables 4, figures 16.
Large Language Models (LLMs) demonstrate strong potential for clinical telemedicine, yet current systems lack structured reasoning, guideline grounding, and decision transparency. We propose DocAgent-XAI, an entropy-guided, multi-agent Retrieval-Augmented Generation (RAG) framework that conducts adaptive, multi-turn clinical interviews. Our approach uses Shannon entropy over a dynamic diagnostic hypothesis distribution to control the progression of a four-level clinical interview: an LLM generates each question from the accumulated clinical state, while entropy, together with information-completeness and confidence criteria, governs when the interview advances between cognitive levels and when it terminates, culminating in guideline-grounded differential diagnoses (DDx). The system provides intrinsic explainability by explicitly tracking hypothesis evolution, entropy-driven decisions, and structured evidence chains. To support scalable and reproducible evaluation, we employ an LLM-based patient simulator grounded in annotated MediTOD profiles, in the spirit of recent simulator-based DDx evaluation [17, 34]. Experiments on the MediTOD dataset (231 respiratory cases), across two LLM backbones and four baselines (single-shot, conversational, RAG, and agentic), show that our method achieves the lowest diagnostic error rate (4.3
