This paper presents a trajectory tracking control method for autonomous vehicles based on Temporal Deep Unrolling-based Model Predictive Control (TDU-MPC). By temporally unrolling the state transitions of the vehicle dynamics to obtain a deep neural network and utilizing backpropagation, the proposed method enables efficient optimization of control inputs subject to complex nonlinearities that challenge conventional approaches. Comprehensive simulation experiments across diverse reference trajectories and disturbance conditions demonstrate that the proposed TDU-MPC consistently outperforms conventional Linear Time-Varying MPC (LTV-MPC), achieving superior tracking accuracy with smaller cumulative lateral error while exhibiting strong robustness to disturbance. Additional experiments using a hand-drawn, bird-shaped trajectory confirm the method’s ability to stably track complex and highly nonlinear trajectories. These findings suggest that TDU-MPC offers a promising framework for achieving high-precision and robust trajectory tracking.

Load frequency control is essential for maintaining power system stability, especially under uncertainties and input delays. This paper proposes a reinforcement learning-based dual-channel dynamic event-triggered fixed-time load frequency control approach for uncertain multi-area power systems with input delays. A non-singular fast terminal sliding mode technique is employed to guarantee that the tracking error converges within a fixed time. To address system uncertainties and input delays, actor neural networks are designed to estimate the modeling uncertainties and provide compensation, and critic neural networks evaluate execution costs. To further enhance efficiency, a dual-channel event-triggered mechanism is designed, reducing communication overhead through independent dynamic event-triggering strategies for control input and output channels. The stability of the proposed method is rigorously analyzed using the Lyapunov method. Simulation results demonstrate faster convergence, reduced communication costs, and improved frequency stability compared to existing methods.

The shepherding problem refers to guiding a group of agents (called sheep) to a specific destination using an external agent with repulsive forces (called shepherd). Although various movement algorithms for the shepherd have been explored in the literature, there is a scarcity of methodologies for selective guidance, which is a key technology for precise swarm control. Therefore, this study investigates the problem of guiding a single target sheep within a swarm to a given destination using a shepherd. We first present our model of the dynamics of sheep agents and the interaction between sheep and shepherd agents. The model is shown to be well-defined with no collision if the interaction magnitude between sheep and shepherd is reasonably limited. Based on the analysis with Lyapunov stability principles, we design a shepherd control law to guide the target sheep to the origin while avoiding collisions among sheep agents. Experimental results demonstrate the effectiveness of the proposed method in guiding the target sheep in both small and large scale swarms.

This work deals with the optimal epidemics policy-seeking problem on networks-of-networks (NoN) in the presence of unknown malicious adding-edge attacks. This problem is investigated in a framework of games-of-games (GoG), in which the conflicts between each network policymaker and the attacker are captured by a series of the Stackelberg games, while all network policymakers together compose a Nash game. First, the tolerable maximum attack magnitude is investigated and given implicitly. Then, we prove the existence of the gestalt Nash equilibrium (GNE) under mild attacks bounded by the above magnitude. A Heuristic algorithm based on iterative geometric programming is proposed to seek the GNE of the above GoG, whose asymptotical convergence is verified. Correspondingly, a greedy Heuristic strategy for the malicious attacker to compromise the NoN topology is developed. The practicability and validity of the above theoretical results and algorithms are illustrated via a simulation example.

An aerial base station (ABS), i.e., unmanned aerial vehicle-mounted base station, has a significant potential to effectively boost the coverage of next-generation wireless networks, having capability of adaptively serving traffic increase in temporary events (i.e., hotspots). However, designing an efficient 3D deployment of ABSs is a considerably complicated problem due to its high degree of freedom and inter-cell interference among ABSs. In this paper, we propose a novel distributed 3D ABS deployment method for providing on-demand downlink communications. To consider the spatial and temporal variations of user locations due to user activities, we model them by an inhomogeneous point process. By analyzing the performance metrics under this model and applying a distributed push-sum, we develop an ABS deployment algorithm with theoretical convergence guarantee that solves the maximization problems of the overall communication quality in a distributed and iterative manner. In our method, each ABS updates its position based on its local information by communicating with its neighboring ABSs. Furthermore, we propose an estimation method of the overall user density from partial observation of ground sensors. Simulation results demonstrate that our method can efficiently improve the overall communication quality and can be applied to a dynamic network.

Deployment of unmanned aerial vehicles (UAVs) performing as flying aerial base stations (BSs) has a great potential of adaptively serving ground users during temporary events, such as major disasters and massive events. However, planning an efficient, dynamic, and 3D deployment of UAVs in adaptation to dynamically and spatially varying ground users is a highly complicated problem due to the complexity in air-to-ground channels and interference among UAVs. In this paper, we propose a novel distributed 3D deployment method for UAVBSs in a downlink network for on-demand coverage. Our method consists mainly of the following two parts: sensing-aided crowd density estimation and distributed push-sum algorithm. The first part estimates the ground user density from its observation through on-ground sensors, thereby allowing us to avoid the computationally intensive process of obtaining the positions of all the ground users. On the basis of the estimated user density, in the second part, each UAV dynamically updates its 3D position in collaboration with its neighboring UAVs for maximizing the total coverage. We prove the convergence of our distributed algorithm by employing a distributed push-sum algorithm framework. Simulation results demonstrate that our method can improve the overall coverage with a limited number of ground sensors. We also demonstrate that our method can be applied to a dynamic network in which the density of ground users varies temporally.