Efficient nitrogen supply is central to sustainable crop production and precision agriculture. This paper develops a distributed nonconvex optimization framework for nitrogen fertilizer management in field crops. A predictive control model with crop nitrogen dynamics is formulated; binary fertilizer on–off decisions make the resulting optimization problem nonconvex. A network-based recurrent neural network (RNN) is constructed, and its convergence to a Karush–Kuhn–Tucker (KKT) point of the formulated model is established. A collaborative neurodynamic mechanism is then used to run multiple distributed RNNs with a population update, so that KKT candidates from different initial states can be compared and refined. In the numerical tests, the computed schedules satisfy the nitrogen constraints and increase yield while reducing nitrogen use. Simulations on a 100-field case and a large-scale 1000-field case over a 30-day horizon verify the proposed framework.
This paper investigates dynamic positioning event-triggered control for unmanned marine vehicle (UMV) systems subject to denial-of-service (DoS) attacks and fading channels. First, the nonlinear UMV system is established as a T-S fuzzy model. Then, an adaptive event-triggered mechanism (ETM) is proposed to optimize system resource utilization. A proper T-S fuzzy extended state observer (ESO) that depends on the DoS attack pattern is constructed to address the unavailability of state information and to estimate unknown disturbances. Based on the estimated information from the ESO, an observer-based fuzzy switching controller is employed to guarantee system performance. Furthermore, sufficient conditions are provided for the mean-square exponential ultimately bounded of the closed-loop system. Finally, a simulation example is presented to verify the effectiveness of the proposed control architecture.
This paper investigates the distributed inverse optimal safety consensus problem for multi-robot systems subject to external disturbances. To address this, a novel hierarchical inverse optimal framework is constructed with two layers. In the upper layer, a distributed optimal trajectory generator is developed, which generates the nominal trajectories for each robot. In the lower layer, an inverse optimal safety controller is presented based on input-to-state safety high-order control barrier functions, including a nominal robust consensus controller and an inverse optimal safety filter. By means of both layers, the multi-robot systems asymptotically achieve infinite-horizon optimal safety consensus by maximizing a meaningful cost functional in the presence of disturbances. Simulations demonstrate the effectiveness of the proposed control strategy.
This paper delves into the demand tracking control problem for a class of coupled re-entrant manufacturing networks (RMNs) utilizing adaptive event-triggered control strategies. The network modeling progresses hierarchically: 1) characterizing the local mass conservation law via explicit dynamical equation at the individual node level; 2) extending to the global mass conservation law for a single RMN; 3) generalizing to a comprehensive global mass balance formulation for a class of coupled RMNs. This multi-scale modeling strategy, transitioning from the local node dynamics to the network-level interactions, yields a rigorous hyperbolic-PDE-based continuum approximation model that efficiently captures the spatiotemporal evolutions of production flows in interconnected manufacturing networks. To reduce unnecessary control actions while maintaining precise regulation of the work-in-progress (WIP) quantity, two novel adaptive event-triggered control schemes incorporating the node-level and the edge-level coordination mechanisms, respectively, are developed for the coupled RMNs. These adaptive triggering mechanisms dynamically adjust triggering thresholds and control gains online, with Lyapunov-based analysis providing sufficient conditions guaranteeing asymptotic tracking of the desired production profiles and exclusion of the Zeno behavior. Numerical simulations validate the effectiveness of the control schemes.
Bipartite consensus is a common phenomenon in multi-agent swarm behavior, and its implementation typically requires signed networks to have a structurally balanced constraint in existing studies. Unlike such studies, this paper investigates multi-agent asynchronous bipartite consensus behavior on quasi-structurally balanced signed networks, which greatly relaxes the existing structurally balanced constraint. A rigorous and comprehensive analysis of asymptotic convergence is conducted by leveraging the infinite product convergence of super-stochastic matrices. In particular, a quasi-structurally balanced condition for achieving asynchronous bipartite consensus is established, along with algebraic inequality constraints on the positive and negative edge weights, which represent cooperative and competitive interactions among agents, respectively. Finally, simulation results, including quantitative comparisons with existing methods, are provided to verify the theoretical results.
Data confidentiality is a key challenge, where cloud computing integrates with control theory, in order to establish cloud-based control systems. As an effective solution to beat this challenge, benefiting from various homomorphic encryption algorithms enables us to ensure data privacy in cloud-based control systems. This paper proposes to use a modified version of the Paillier encryption algorithm in implementation of cloud-based control systems, which allows encryption with rational numbers (instead of the special class of integer numbers). Furthermore, a novel representation for dynamic controllers possessing irrational parameters in the type of algebraic numbers is introduced, in order to be encrypted by the aforementioned encryption algorithm in an innovative framework. As a comparative advantage, the presented framework reduces the need for quantization and scale factor selection, which have their own difficulties in implementation of cloud-based control systems. Two numerical examples are provided to demonstrate the usefulness of the proposed framework in running away from the above-mentioned difficulties in implementation of encrypted controllers.
This article investigates the fully distributed prescribed-time(PT) full-state consensus problem of the multiagent systems (MASs) under the communication link faults(CLFs). Unlike the existing related results, the developed scheme not only solves the MASs' full-state PT control problem, but also tolerates the existence of CLFs and even allows the CLFs are discontinuous. First, a fully distributed PT observer is designed for each follower, enabling the reconstruction of the leader's state within a prescribed time under CLFs. By using the adaptive method to estimate the eigenvalues of the unknown topology matrices under CLFs, and integrate this with the infinite gain characteristic of the PT function, the adverse effects of CLFs on the observer design are effectively mitigated. Then with the help of the observer, by the time-varying gain, a PT controller is proposed for the follower by which the full-state consensus tracking for the MASs is achieved in a prescribed time. Ultimately, the performance of the control protocol is verified via simulation.
This article delves into the $H\_{\infty }$ control issue within the context for noisy sampled-data networked control systems, specifically addressing the challenges posed by dual-channel false data injection (FDI) attacks in conjunction with packet losses, all while accounting for the complexities introduced by noisy sampled-data intervals. A unified feedback interconnection model with an exponential decay rate is established. The novel estimates of the upper bound on the expectation of a random variable are derived for the stochastic characteristics of noisy sampled-data intervals and dual-channel packet dropouts. Subsequently, an operator that satisfies the definition of stochastic discrete-time integral quadratic constraint is given, which combines the dual-channel false data injection attacks signal upper bound and the $H\_{\infty }$ performance index. In particular, an exponential stability condition with an exponential decay rate is provided. Based on this, the controller that can ensure the exponential stability of the networked control systems is designed and the corresponding controller algorithm is proposed. Finally, an aircraft longitudinal motion model is employed as an illustrative example to verify the effectiveness of the proposed algorithm.
Networked systems confront rapidly growing security threats. This paper proposes a deceptive defense scheme in these networked systems modeled by attack graphs. Our deceptive approach involves modeling these systems using a deceptive game framework which accounts for the impact of deception on security resource allocation by the defender and the respective response of the attacker under such a defender's deceptive investment. In our framework, the defender seeks to deceive the attacker by providing misconceptions about the distribution of security resources on the attack graph's edges to the attacker, leading the attacker to choose wrong attack paths (that are more secure for the defender). We first establish the existence of equilibrium and provide an illustrative example to highlight the effects of the deception framework. Subsequently, we provide mathematical analysis of two proposed deceptive schemes to minimize the total expected cost of the defender. We then analyze the gain derived from this deception by the defender, where we adapt a widely recognized metric to quantify the extent of this gain. We also provide bounds for such gain and show its exponential growth with the increase of the defender's security budget. To assess our models, we evaluate our deception schemes using two representative real-world networked systems and compare the different deception setups and the resulting gain in the system's security level for different investment strategies.
In this paper, the novel identification criteria for weighted matrices (WMs) of delayed multi-layer stochastic complex network with Lévy noise (DMSCNL) are investigated based on adaptive synchronization. Compared with previous work, this is the first study on the problem of identification for WMs of stochastic complex network with Lévy noise. Specifically, considering DMSCNL with unknown WMs as the original system, an auxiliary system with an adaptive feedback controller (AFC) and estimated weighted values is designed to track the original system by combining graph theory with the non-negative semi-martingale convergence theorem. Furthermore, the estimated weighted values in the auxiliary system can converge to the unknown weighted values in the original system under an excitation condition. That is, the corresponding identification criteria for WMs of the original system are derived. Moreover, the theoretical results are validated through a two-layer recurrent neural network. Finally, numerical simulations well verify the validity of theoretical results.
This paper investigates an aggregative game with a shared linear coupling inequality constraint and local compact convex set constraints, where agents are distributed across a communication network. The goal is to find a generalized Nash equilibrium (GNE) of the game in a distributed manner over time-varying unbalanced graphs. To this end, we propose a distributed discrete-time GNE seeking algorithm that integrates the push-sum protocol with an aggregative information tracking mechanism. The algorithm also accommodates nonidentical step-sizes. By reformulating the algorithm as an extended Krasnosel'skiMann iteration, we establish its convergence to a variational GNE and analyze its convergence rate. Numerical simulations on a generalized Nash-Cournot game validate the effectiveness of the proposed algorithm.
We study a multi-agent system governed by a state-dependent discontinuous vector field arising from closest-target selection dynamics. Unlike conventional approaches that regularize the target-assignment rule, we directly incorporate the closest-target selection rule into the continuous-time dynamics and analyze the resulting discontinuities within Filippov's differential inclusion framework. The main analytical challenge stems from the nonsmooth relaxation force induced by target switching, which prevents the direct application of classical Lyapunov methods. To address this difficulty, we construct a Lyapunov functional compatible with Filippov set-valued derivatives and apply a LaSalle-type invariance principle for differential inclusions. A key step of the analysis is the explicit characterization of the Filippov set-valued relaxation force as the convex hull of admissible closest-target directions. Under verifiable conditions on the initial data, system parameters, and kernel functions, we establish global existence and asymptotic convergence of all Filippov solutions. In particular, we show that each tracking agent converges to a point in a finite set consisting of points that lie in the convex hull of the targets closest to those points, while all velocities decay to zero. This finding provides the strongest asymptotic convergence result that can be expected for discontinuous closest-target selection dynamics. The theoretical results are further illustrated by numerical simulations highlighting geometric obstructions in multi-target configurations.
This paper studies the distributed Nash equilibrium (NE) seeking problem for constrained multi-cluster aggregative games over digraphs, where the closed convex set constraints are involved. In the considered game, the players within the same cluster collaboratively minimize the collective cost within locally closed convex sets regardless of the interests of other clusters, where the local objective functions depend not only on local decisions but also on a global aggregative variable. A distributed algorithm is designed to seek the NE based on distributed aggregate estimation and gradient tracking. Specially, the eigenvector learning method is employed to address the information asymmetry caused by unbalanced digraphs. Moreover, the method of feasible direction is incorporated into the proposed algorithm to handle the involved closed convex set constraints. Under certain conditions, it is proved that the proposed algorithm achieves a linear convergence rate. Finally, the proposed algorithm is applied in a numerical simulation associated with the Energy Internet System to verify its effectiveness.
We consider the aggregation of distributed energy resources (DERs), such as solar PV, energy storage, and flexible loads, by a profit-seeking aggregator participating directly in the wholesale market under distribution network access constraints. We propose a competitive DER aggregator (DERA) model that directly controls local DERs to maximize its profits, while ensuring each aggregated customer gains a surplus higher than their surplus under the regulated retail tariff. The DERA participates in the wholesale electricity market as virtual storage with optimized generation offers and consumption bids derived from the proposed competitive aggregation model. Also derived are DERA's bid curves for the distribution network access and DERA's profitability when competing with the regulated retail tariff. We show that, with the same distribution network access, the proposed DERA's wholesale market participation achieves the same welfare-maximizing outcome as when its customers participate directly in the wholesale market. Extensive numerical studies compare the proposed DERA with existing methods in terms of customer surplus and DERA profit. We empirically evaluate how many DERAs can survive in the competition at the long-run equilibrium, and assess the impacts of DER adoption levels and distribution network access on short-run operations.
In this paper, the finite-time distributed time-varying optimization problems for high-order agents are investigated, where the gradient derivative of time-varying cost functions is unknown. First, a finite-time distributed optimal trajectory estimator which do not rely on the Hessian information and the partial derivative of gradient with respect to time is constructed. Then, based on the generated optimal trajectory estimation signal, a distributed time-varying optimization algorithm for high-order agents with globally practically finite-time convergence is designed, where the bound of optimization error can be arbitrarily small by regulating controller parameters. Moreover, the rigorous convergence analysis is provided as well. In final, a simulation example is conducted to validate the proposed theoretical results.
This paper addresses distributed optimization problems with compact convex set constraint and nonsmooth objectives over weight-unbalanced directed communication graphs, where the objective function is a sum of local convex functions endowed only by the corresponding agent. The weight unbalance destroys the doubly-stochastic property required by standard consensus algorithms, while the nonsmooth objectives and coupled constraints further complicate distributed computation. To tackle these challenges, we propose a novel continuous-time projection algorithm that achieves finite-time weight balancing and finite-time convergence to the feasible set from any initial values, followed by asymptotic convergence to an optimal solution. Finally, two numerical examples are performed to substantiate the effectiveness, demonstrating its strong adaptability to both weight-balanced and weight-unbalanced graphs compared with existing methods.
In this paper, we tackle the problem of computing the stationary distribution vector of a Markov Chain that is dense but whose structure has several entries equal to the same value. We first decompose the Markov matrix such that it exhibits a sparse structure (hence the quasi-sparse terminology) and provide its formulation as the solution of an optimization problem, a linear equation or the solution of an eigenvector problem. A result is presented that shows optimization algorithms to grow quadratically the problem condition number. The solution as an eigenvector problem using the Power iteration corresponds the Jacobi method when applied to the corresponding equation. Thus, we propose the use of the Gauss-Seidel or the Successive Over-Relaxation (SOR) methods to tackle the problem using synchronous and asynchronous communication that is shown to have better worst-case convergence rate. We illustrate through simulations for scale-free Markov Chains, based on the Barabási-Albert model, that the proposed algorithms achieve faster convergence in particular when subsequent solutions are required with Markov chains that are similar to the previous computation.
Although encrypted control systems ensure confidentiality of private data, it is challenging to detect anomalies without the secret key as all signals remain encrypted. To address this issue, we propose a homomorphic encryption scheme for dynamic controllers that automatically discloses the residue signal for anomaly detection, while keeping all other signals private. To this end, we characterize the zero-dynamics of an encrypted dynamic system defined over a finite field of integers and incorporate it into a Learning With Errors (LWE) based encryption scheme. We then present a method to further utilize the disclosed residue signal for implementing dynamic controllers over encrypted data, without requiring re-encryption even when they have non-integer state matrices.
