Publications

AerialMind: Towards Referring Multi-Object Tracking in UAV Scenarios

Referring Multi-Object Tracking (RMOT) aims to achieve precise object detection and tracking through natural language instructions, representing a fundamental capability for intelligent robotic systems. However, current RMOT research remains mostly confined to ground-level scenarios, which constrains their ability to capture broad-scale scene contexts and perform comprehensive tracking and path planning. In contrast, Unmanned Aerial Vehicles (UAVs) leverage their expansive aerial perspectives and superior maneuverability to enable wide-area surveillance. Moreover, UAVs have emerged as critical platforms for Embodied Intelligence, which has given rise to an unprecedented demand for intelligent aerial systems capable of natural language interaction. To this end, we introduce AerialMind, the first large-scale RMOT benchmark in UAV scenarios, which aims to bridge this research gap. To facilitate its construction, we develop an innovative semi-automated collaborative agent-based labeling assistant (COALA) framework that significantly reduces labor costs while maintaining annotation quality. Furthermore, we propose HawkEyeTrack (HETrack), a novel method that collaboratively enhances vision-language representation learning and improves the perception of UAV scenarios. Comprehensive experiments validated the challenging nature of our dataset and the effectiveness of our method.

Chat with ChatGPT on Industry 5.0: Learning and Decision-Making for Intelligent Industries

The current ChatGPT phenomenon has signaled a new era of Artificial Intelligence moving from Algorithmic Intelligence to Linguistic Intelligence where interactive activities between actual and artificial, real and virtual, human and machine play an active and important role online and in real-time. At IEEE/CAA JAS, we are interested in investigating the impact and significance of this new era on industrial development, especially control and automation for manufacturing and production.

Fixed-time pinning-controlled synchronization for coupled delayed neural networks with discontinuous activations

This paper deals with the fixed-time synchronization problem of coupled delayed neural networks with discontinuous activations. Based on pinning control, a discontinuous controller is firstly proposed to guarantee that coupled neural networks achieve synchronization with a desired trajectory in finite time. Then, a discontinuous fixed-time controller is designed. With the fixed-time controller, the settling time can be estimated regardless of initial conditions. By providing a topology-dependent Lyapunov function, some criteria of finite-/fixed-time synchronization are derived. Finally, two numerical examples are given to show the effectiveness of the proposed controllers.

Novel Leader–Follower-Based Particle Swarm Optimizer Inspired by Multiagent Systems: Algorithm, Experiments, and Applications

In this article, as inspired by multiagent systems, a novel leader–follower-based particle swarm optimization (LFPSO) algorithm is presented where the particles are classified into leaders and followers according to their respective roles. The leaders are responsible for searching a wide range of the optimal candidate solutions so as to ensure the diversity of the particle population, and the followers are dedicated to seeking the global-best solution in order to guarantee the convergence of particles. A controller parameter is introduced to fine tune the impact of the leaders on the followers. Owing to the leader–follower mechanism, the proposed LFPSO algorithm not only maintains the diversity of the particle population but also improves the possibility of escaping from the locally optimal solution. It is demonstrated via experimental results that the proposed LFPSO algorithm significantly improves the accuracy and convergence rate of conventional particle swarm optimization algorithms. Furthermore, the LFPSO algorithm is successfully applied to denoise real-time signals in oilfield pipeline network and its superiority over existing denoising algorithms is verified as well.

A brief survey of recent results on control and filtering for networked systems

During the past few decades, networked systems have gained increasing attention and a wider range of applications have been found in areas such as industrial control and signal processing. This paper provides a brief survey of recent results on control and filtering for networked systems. In-depth analysis and discussion is first made on networked control and event-triggered control, where existing research methods on networked control systems are summarized. Then, network-based filtering is also reviewed. Finally, some challenging problems are presented to direct future research.

On supervised learning of sliding observer

A sliding observer for a high-order nonlinear uncertain system is examined from the machine learning perspective. The supervised learning mechanism embedded in the sliding observer system, first introduced by Slotine in 1987, is explored in details by integrating a master-slave learning concept between the subsystems of the sliding observer. In particular, the master system acts as a coordinator that coordinates the slave systems to achieve finite-time convergence. During the learning process, the slave systems are forced to learn the finite error convergence properties of the closed-loop error dynamics of the master system by injecting the switching component in all subsystems. Finite-time error convergence of the sliding observer is proved by adopting the idea of modulation transfer function into the closed-loop error dynamics. The simulations are conducted to verify the theoretical analysis.

Output-based event-triggered ℋ<inf>∞</inf> control for sampled-data control systems with nonuniform sampling

This paper is concerned with the event-triggered output feedback ℋ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> control for sampled-data control systems with nonuniform sampling. Firstly, an output-based event-triggered communication scheme is proposed, in which the state of the system is sampled with a nonuniform sampling period. Under this scheme, whether or not the sampled data should be transmitted is determined by a predetermined output-based event-triggering condition. Secondly, an event-triggered networked control system is modeled as a time-delay system by taking into consideration the output-error between the output at the current sampling instant and the output at the last transmitted sampling instant. Thirdly, stability and stabilization criteria are derived to guarantee the uniform ultimate bounded stability and the desired performance while using less communication resources. Finally, two illustrative examples are used to show the effectiveness of the proposed method.

ESO-Based Actor-Critic Structure for Optimal Velocity Control of an Autonomous Surface Vehicle with Partial Unknown Model Nonlinearities

This paper investigates the surge velocity control of an autonomous surface vehicle (ASV) in the context of a specific performance index subject to partially un

Algorithm Design and Regret Analysis for Aggregative Optimization in Open Multi-Agent Systems

OSE: On-Site State of Health Estimation for Li-ion Battery using Real-Time Field Data

On-site lithium-ion battery state of health (SoH) estimation is of crucial importance for reliable operations of electric vehicles (EVs). Yet, due to the low-quality of unlabeled real-time field data, diverse operating environments of in-service EVs, and limited computational capability of onboard devices, existing techniques established on data from well-controlled experimental environments are not practical for real world EVs' SoH estimation. Accurate and rapid SoH estimation based on field data of in-service EVs still remains quite challenging. To tackle this challenge, we present an on-site SoH estimation (OSE) method using in-service EV field data through a new knowledge embedded deep transfer learning (DTL) model. Initially, an universal data preprocessing approach integrating mechanism knowledge is designed to process low-quality data under diverse operating environments. Then, we develop a domain adaptive hybrid deep neural network (DAHDNN) model suitable for unlabeled field data, which can be deployed via edge cloud collaborative framework to meet actual computational capability. We demonstrate the superiority of our method across four real datasets, where, OSE's estimation error is decreased by up to 78.5% compared with the state-of-the-art methods. The results indicate that the proposed method has good generalizability and reliability for SoH estimation on real-time field data.

Securing the Future After PagerBombs: Lifecycle Protection of Smart Devices via Blockchain Intelligence

The Lebanese wireless device explosion incident has drawn widespread attention, involving devices such as pagers, walkie-talkies, and other common devices [1].

Investigating the effects of time-delays on stochastic stability and designing l1-gain controllers for positive discrete-time Markov jump linear systems with time-delay

This paper is concerned with stochastic stability, l 1-gain performance analysis and positivity-preserving l 1-gain controller design for a positive discrete-time Markov jump linear system with time-delay. First, necessary and sufficient conditions for stochastic stability of the system are derived by constructing a linear co-positive stochastic Lyapunov functional and establishing a system equation whose state variables consist of the mathematical expectation of the markovianized states and whose coefficient matrices depend on time-delay and the transition probability. It is revealed that stochastic stability of the positive discrete-time Markov jump linear system with time-delay is influenced by the size of time-delay and it is demonstrated by an example that the effect of time-delay on stochastic stability can be either positive or negative. Second, exact computation on an l 1-gain index of a stochastically stable positive discrete-time Markov jump linear system with time-delay is presented, and a necessary and sufficient condition for the l 1-gain performance is derived in the form of linear programming. Third, an iterative algorithm is proposed to design a positivity-preserving l 1-gain controller, and in single-input case, an optimal controller is obtained analytically such that the closed-loop system achieves the minimal l 1-gain performance. Then, a modified pest’s structured population dynamic model is developed to illustrate the effectiveness of the designed method.

On the Isometric Embedding of Torus in $R3$

International Press of Boston - publishers of scholarly mathematical and scientific journals and books

A new unidirectional coupling for global synchronization of two Hindmarsh-Rose neurons

In this paper, a new unidirectional coupling for global synchronization of two Hindmarsh-Rose neurons is designed. Compared with existing results, external control for global synchronization is no longer required. A synchronization criterion of two Hindmarsh-Rose neurons with rigorous mathematical proof is derived. An example is given to illustrate the effectiveness of the obtained result.

Delayed non-fragile H∞ control for offshore steel jacket platforms

This paper is concerned with a delayed non-fragile [Formula: see text] control scheme for an offshore steel jacket platform subject to self-excited nonlinear hydrodynamic force and external disturbance. By intentionally introducing a time-delay into the control channel, a delayed robust non-fragile [Formula: see text] controller is designed to reduce the vibration amplitudes of the offshore platform. The positive effects of the time delays on the non-fragile [Formula: see text] control for the offshore platform are investigated. It is shown through simulation results that (i) the proposed delayed non-fragile [Formula: see text] controller is effective to attenuate the vibration of the offshore platform; (ii) the control force required by the delayed non-fragile [Formula: see text] controller is smaller than that required by the delay-free non-fragile [Formula: see text] controller; (iii) the time delays can be used to improve the control performance of the offshore platform.

Distributed <i>H</i> _∞ filtering over sensor networks with heterogeneous Markovian coupling intercommunication delays

This study is concerned with distributed H∞ filtering for continuous-time linear systems over sensor networks with heterogeneous Markovian coupling intercommunication delays. The set of sensor nodes forms a sensing and communication network whose topology is modelled by a directed graph that describes the measurement exchanged among neighbouring sensor nodes. Heterogeneous random coupling delays modelled by a Markov process are considered in the intercommunication between interacting sensor nodes. A refined two-step decoupling technique is presented to deal with the complicated coupling of exchanged measurement in the presence of the delays. A sufficient condition on the existence of desired distributed H∞ filters is derived such that the resultant filtering error system is mean square exponentially stable with prescribed weighting average H∞ performance. Two illustrative examples are given to show the effectiveness of the proposed results.

Distributed Gradient Tracking for Differentially Private Multi-Agent Optimization With a Dynamic Event-Triggered Mechanism

Distributed optimization achieves a minimized objective function through collaboration among distributed agents. Considering limited communication capabilities and privacy concerns, this article proposes a dynamic event-triggered differentially private gradient-tracking algorithm for distributed optimization. The communication requirement is reduced by event triggering, while the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\epsilon$</tex-math> </inline-formula> -differential privacy is guaranteed by perturbations on states and the tracking of the average gradient. The convergence point is uniquely determined by the noise injected to the tracking. Sufficient conditions for stepsizes are established theoretically to guarantee the convergence in mean and almost surely. Moreover, the theoretical privacy level is rigorously obtained and the positive effect of the event-triggered communication on the privacy is also discussed. Simulations are conducted for the classification of the dataset on the stability of a 4-node star power system to verify the theoretical findings.

Repetitive Control: Basic Concept, Fundamental Theory, and Practical Applications