Visual storytelling systems struggle to maintain character identity across frames and link actions to appropriate subjects, frequently leading to referential hallucinations. We address these issues through entity grounding and propose StoryReasoning, a dataset containing 4178 stories derived from 52,016 movie images, with both structured scene analyses and grounded stories. Each story maintains character and object consistency across frames while explicitly modeling multi-frame relationships through structured tabular representations. Our approach features cross-frame object re-identification using visual similarity and face recognition, chain-of-thought reasoning for narrative modeling, and a grounding scheme linking textual elements to visual entities across frames. We fine-tune Qwen2.5-VL 7B, creating Qwen Storyteller, which performs end-to-end object detection, re-identification, and landmark detection while maintaining consistent object references throughout the story. Under a panel of Large Language Model (LLM) judges, evaluation shows a reduction from 4.06 to 3.56 (−12.3%) hallucinations per story and improved creativity from 2.58 to 3.38 (+31.0%) compared to the base model. Larger closed-source models remain ahead on free-form narrative quality; the contribution of our approach is explicit structured grounding with entity identifiers that persist across frames.
This study presents an experimental investigation of the behaviour of a steel seismic shelter installed on the top floor of a vulnerable unreinforced masonry (URM) building during its collapse. A 1:2 scale, four-storey stone-masonry building model with timber floors and deliberately weak wall-to-floor connections, accommodating a 1:2 scale shelter, was tested to collapse on the 3D shaking table of LNEC within the SHELTER project (Structural Hyper-resisting Element for Life-Threatening Earthquake Risk). The collapse developed in two phases, involving the failure of the top-floor connection, the rotation of the shelter and its subsequent descent onto the lower floors. Given the destructive nature of the test, the kinematics of the building and of the shelter were monitored by high-speed and action cameras tracking optical and infrared targets, complemented by wired and wireless high-capacity accelerometers mounted on the shelter. In the second phase of the collapse, the shelter moved approximately 3.0 m vertically and 1.0 m horizontally. Peak impact accelerations exceeded 50 g at the shelter base plate, whereas those recorded along the shelter columns were substantially lower. The mean downward acceleration of the shelter during its main descent was approximately 0.58 g, slightly above the value of 0.5 g assumed in design. The results show that the rotation of the shelter generates impacts in directions not considered in the original design of the safety seats, and highlight the importance of a firm shelter-to-floor anchorage and of multi-directional energy-absorbing seats for occupant protection. The limitations associated with a single, distorted-scale test are discussed.
Photovoltaic (PV) power output is significantly influenced by environmental and climatic variables, such as the sun's irradiance, air humidity, and ambient temperature. In addition, most PV power sources have relatively small output voltages not suited for direct utilization by common loads. Additionally, PV power output depends also on the PV operating voltage and current. In this sense, a maximum power point tracking (MPPT) algorithm is required to achieve maximum power output and improved efficiency in PV systems. Due to shadowing or fast irradiance changes, MPPT algorithms must operate at high speed and accuracy. Thus, this paper proposes a MPPT control strategy based on incremental resistance (INR) combined with a proportional-integral (PI) compensator (INR-PI). This approach modifies the seldom used INR MPPT algorithm to set the current and combines it with a proportional-integral compensator to regulate the I-V operating point of the PV panel. This reformulates the INR principle as a PV panel voltage regulation problem ensuring closed-loop stability and disturbance rejection. The theoretical analysis considers the dynamics of the input current in a quadratic Buck-Boost converter with an extended duty-cycle range in the Buck voltage region, connected to a DC microgrid. The proposed MPPT technique is designed to ensure fast and efficient operation of PV systems under rapidly changing environmental conditions. To validate this control system, initial simulations were conducted using MATLAB/Simulink, followed by experimental testing with a laboratory prototype setup. Both numerical and experimental results demonstrate the effectiveness of the feedback-based incremental resistance algorithm in tracking the maximum power point (MPP) of PV systems when irradiance fluctuates quickly.
Uncertainty in future demand poses a critical challenge for decision-making, particularly in strategic Emergency Medical Services (EMS) logistics planning and public service resource allocation, as it directly affects decision outcomes. For example, in EMS systems, failing to properly account for demand uncertainty during planning may lead to suboptimal capacity decisions, resulting in insufficient or inefficient resource allocation. To address this challenge, we propose a novel hybrid scenario generation method, called Residual Scenario Generation (RSG), which constructs representative uncertainty scenarios from forecasting residuals while maintaining a compact scenario set. The method is evaluated in the context of strategic-tactical EMS capacity planning using a bi-objective two-stage stochastic programming model that considers ambulance base location and fleet sizing decisions under demand uncertainty, with personnel requirements implicitly determined by the crew composition associated with each vehicle type. The model considers a multi-period planning horizon in which long-term configuration decisions are evaluated while ensuring that long-term demand can be adequately satisfied. Using real data provided by the Portuguese EMS provider (INEM), we compare the RSG method with Monte Carlo, Bootstrap Sampling, and Latin Hypercube Sampling. Our results show that the different scenario generation methods produce comparable planning outcomes, while RSG achieves substantially lower computational effort and stable performance using a considerably smaller number of scenarios. These characteristics make RSG particularly suitable for stochastic planning settings requiring repeated optimization across multiple configurations or sensitivity analyses. Managerial insights are also provided, allowing for better informed decision-making.
The present investigation focuses on rehabilitating cracked subsea pipelines by incorporating a functionally graded material (FGM) liner. Seawater sweeps into the cracked area of the pipeline, generating water pressure at the outer surface of a liner, which leads to “single-lobe” or “double-lobe” deflections because of the constraints of the pipeline. In addition, in practical engineering, issues including inadequate expansion and deflection of the pipeline may cause the liner to become oval. Accordingly, a systematic theoretical scheme is developed to elucidate the buckling characteristics of an encased oval FGM liner under uniform pressure. The critical buckling pressure is computed explicitly from the thin-walled shell principle and energy method. A two-dimensional (2D) simulated model is used to obtain the buckling pressure and pressure-displacement equilibrium path. Theoretical solutions are in close alignment with the other closed-form solutions, simulated results, and experimental results. Finally, the influence of volume fraction exponent and ovality on the critical buckling pressure and critical circumferential force is discussed comprehensively. The conclusions illustrate that the increasing volume fraction exponent and ovality significantly decrease the pressure capacity. As the double-lobe deflection occurs, the liner exhibits a higher pressure capacity than the single-lobe case.
