Protecting buildings—especially critical infrastructure—has become increasingly urgent amid evolving threats. Traditional assessments, based on visual inspections and expert analyses, struggle with the complexity and scale of modern structures. Meanwhile, although object detection is widely used in real-time surveillance, its application to static architectural layouts remains limited. This paper presents a novel use of the YOLOv8m framework to automatically identify structural elements in architectural plans, laying the groundwork for building topological graphs that map potential adversarial attack paths. Our work contributes by creating a curated dataset of real and synthetic building layouts with annotated architectural symbols, designing a customized YOLOv8m pipeline for effective object detection, and developing an initial framework for converting detections into graph-based security models. By integrating deep learning with traditional security methods, this research enhances the modeling and assessment of physical security.
This paper proposes a stabilization method for second-order switched systems composed of stable subsystems. The approach introduces a technique for determining a quadratic Lyapunov function for linear, fractional-order, and integer-order switched systems, to derive an appropriate switching function. The proposed method ensures system stability by designing a suitable switching function, making it applicable to both integer- and fractional-order systems. The results demonstrate the effectiveness of the approach in maintaining stability across different system orders.
A control problem is considered for the dynamic model of network of beams. The beam model is selected to be of Timoshenko type. Assume the turnpike property for the network control problem holds, then the design of the network can be performed for the static model of beams. Therefore, the topological derivative method can be used for optimum design of network. We present an example of control problem and of the optimum design for the network of Timoshenko beams. The cost of control problem serves as a shape functional for the optimum design of network.
Inventory management policies designed using the control theory methods have proven useful and practical over the years. However, due to their specific characteristics, they rarely focus on the holding costs optimization. In this paper a dead-beat controlled discrete-time inventory model is analyzed assuming uniformly distributed demand. The presented method of analyzing the system output distribution is meant to be used as a supplement to the order policies presented in literature, that originate from control theory. It is shown, that the stock level depends directly on the historical demand and therefore its probability distribution can be determined. Knowing this probability, it is possible to determine how much decreasing the reference stock level will affect the inventory system performance and to find a compromise between the holding costs and the probability of realizing all of the orders in time. The obtained results are verified in simulation.
Necessary and sufficient conditions for the reduction of the descriptor (singular) Roesser model to the equivalent standard one are established. Procedure for computation of the equivalent standard Roesser model is given and illustrated by numerical example.
