Keywords: Algebraic and geometric methods for network control, Distributed optimization, Decentralized control and large-scale systems
Abstract: In the large number of layers limit, deep learning architectures like residual networks and transformers naturally converge to controlled ordinary differential equations. Embracing this perspective, the talk takes a geometric control approach to learning and model tuning. Building on new results about the geometry of the space of model parameters that memorize a dataset, we propose new algorithms for these tasks and illustrate their performance. We conclude with open questions bridging geometric control theory, optimization landscapes, and the implicit biases of gradient-based learning.

Keywords: Applications of consensus and gossip algorithms, Sensor networks and distributed processing
Abstract: In this abstract, we present the main results of (Bayram et al., 2023). We study the weighted average consensus problem for a gossip network of agents with vector-valued states. For a given matrix weighted graph, the gossip process is described by a sequence of pairs of adjacent agents communicating and updating their states based on the edge matrix weight. Our key contribution is providing conditions for the convergence of this non-homogeneous Markov process as well as the characterization of its limit set. To this end, we introduce the notion of ``w-holonomy'' of a set of stochastic matrices, which enables the characterization of sequences of gossiping pairs resulting in reaching a desired consensus in a decentralized manner. Stated otherwise, our result characterizes the limiting behavior of infinite products of (non-commuting, possibly with absorbing states) stochastic matrices.

Keywords: Biological networks and epidemics dynamics, Multi-agent systems, Social and economic networks
Abstract: Transmission Neural Networks (TransNNs) introduced by Gao and Caines (2022) connect virus spread models over networks and neural networks with tuneable activation functions. This paper presents the approximation technique and the underlying assumptions employed by TransNNs in relation to the corresponding Markovian Susceptible-Infected-Susceptible (SIS) model with 2^n states, where n is the number of nodes in the network. The underlying infection paths are assumed to be stochastic with heterogeneous and time-varying transmission probabilities. We obtain the conditional probability of infection in the stochastic 2^n-state SIS epidemic model corresponding to each state configuration under mild assumptions, which enables control solutions based on Markov decision processes (MDP). Finally, MDP control with 2^n-state SIS epidemic models and optimal control with TransNNs are compared in terms of mitigating virus spread over networks through vaccination, and it is shown that TranNNs enable the generation of control laws with significant computational savings, albeit with more conservative control actions.

Keywords: Control under communication constraints
Abstract: This paper investigates the Linear Quadratic Regulator (LQR) problem over an unknown Bernoulli packet drop channel. The unknown packet drop probability is estimated using finite samples, then the estimated probability is used to design a formally equivalent optimal controller. If the estimation error is too large, the estimated controller cannot mean-square stabilize the system. So the upper bound on the estimation error is provided to guarantee the stability of the closed-loop system. And we present an analytical expression for the gap between the performance of the estimated controller and the optimal performance. Furthermore, we derive a lower bound of the sample size required for the stabilizability of the estimated controller. Finally, Numerical examples are used to demonstrate our results.

Keywords: Cooperative and distributed learning, Distributed optimization, Multi-agent systems
Abstract: In this abstract we present the main results of Ren et al. (2025). We address distributed learning problems, both nonconvex and convex, over undirected networks. To this end, we design a novel algorithm based on the distributed Alternating Direction Method of Multipliers (ADMM) to address the challenges of high communication costs, and large datasets. Our design tackles these challenges by, i) enabling the agents to perform multiple local training steps between each round of communications; and ii) by allowing the agents to employ stochastic gradients during local computations. We show that the proposed algorithm converges to a neighborhood of a stationary point, for nonconvex problems, and of an optimal point, for convex problems. We also propose a variant of the algorithm to incorporate variance reduction and thereby achieve exact convergence. We show that the resulting algorithm indeed converges to a stationary (or optimal) point, and moreover that local training accelerates convergence. We thoroughly compare the proposed algorithms with the state of the art, both theoretically and through numerical results.

Keywords: Distributed optimization, Applications of consensus and gossip algorithms, Cooperative and distributed learning
Abstract: This paper addresses the problem of computing the sample variance of datasets scattered across a network of interconnected agents. A general procedure is outlined to allow the agents to reach consensus on the variance of their local data, which involves two cascaded (dynamic) average consensus protocols. Our implementation of the procedure exploits the distributed ADMM, yielding a distributed protocol that does not involve the sharing of any local, private data nor any coordination of a central authority; the algorithm is proved to be convergent with linear rate and null steady-state error. The proposed distributed variance estimation scheme is then leveraged to tune personalization in "personalized learning" where agents aim at training a local model tailored to their own data, while still benefiting from the cooperation with other agents to enhance the models' generalization power. The degree to which an agent tailors its local model depends on the diversity of the local datasets, and we propose to use the variance to tune personalization. Numerical simulations test the proposed approach in a classification task of handwritten digits, drawn from the EMNIST dataset, showing the better performance of variance-tuned personalization over non-personalized training.

Keywords: Distributed optimization, Applications of consensus and gossip algorithms, Cooperative and distributed learning
Abstract: Recent technological advances have enabled the deployment of multi-agent systems for a wide range of applications, including the cooperative solutions of learning problems. Therefore, there is a need to design algorithmic methods to enable distributed learning, while guaranteeing their robustness to the practical challenges they may face. In traditional distributed learning approaches, the agents participating in the training process are assumed fixed. However, in practice heterogeneous resources (e.g. battery levels) might result in agents continuously joining and leaving the training. In our contributions we focus on providing a novel algorithm, based on Alternating Direction Method of Multipliers (ADMM), to enable learning over these open networks. Additionally, we analyze the robustness of this algorithm to a broad range of practical challenges besides the openness of the network.

Keywords: Game theory and network games
Abstract: Modern buildings must prioritize occupants' thermal comfort to enhance well-being and productivity. Since individuals have diverse temperature preferences, understanding how interactive thermal tolerance affects optimal indoor conditions is key to balancing comfort and energy efficiency. We propose a Cyber-Physical-Social Systems framework to model interactions between temperature control, building environments, and occupant behavior. Using game theory, we optimize feedback mechanisms and derive Nash equilibrium for thermal tolerance, proving the optimal temperature's bounds exhibit Lipschitz continuity. A parallel best-response algorithm achieves globally optimal costs, maximizing collective comfort. Simulations validate our method's effectiveness.

Keywords: Game theory and network games, Decentralized control and large-scale systems, Graph-theoretic methods for control
Abstract: We consider dynamic games with large subpopulations distributed over large sparse graphs. Each agent on one hand has mean field coupling with all agents located within the same cluster and on the other hand receives impact from neighboring clusters via a graph Laplacian. We aim to derive tractable limit models when the sparse network size tends to infinity, which leads to higher order mean field interaction dynamics. This is justified by asymptotic analysis of the graph Laplacian operator within the large graph limits. Subsequently, the solution equation system of the Laplexion mean field game is obtained within the setting of infinite population and infinite network size.

Keywords: Game theory and network games, Traffic networks
Abstract: Traditional non-atomic selfish routing games present some limitations in properly modeling road traffic. This work introduces a novel type of non-atomic selfish routing game leveraging concepts from the cell transmission model (CTM) by Daganzo (1994). Each network link is characterized by a supply and demand mechanism that enforces capacity constraints based on current density, providing a more accurate representation of real-world traffic phenomena. We characterize the Wardrop equilibria and social optima of this game and identify a previously unrecognized inefficiency in selfish routing: partially transferring Wardrop equilibria, where only part of the exogenous flow goes through the network.

Keywords: Multi-agent systems, Applications of consensus and gossip algorithms, Distributed constrained control and MPC
Abstract: This work shows an approach to achieve output consensus among heterogeneous agents in a multi-agent environment where each agent is an uncertain linear system subject to input and state constraints. The communication among agents is described by a time-varying directed graph with bounded communication delays. The approach relies on achieving consensus among the references of the agents and that each agent has a Model Predictive controller (MPC) that tracks piecewise constant references asymptotically. Since only the reference contains diffusive terms from its neighbors, the approach has a larger feasible domain than approaches that use diffusive terms in the control. Examples are provided to illustrate the advantages of the approach.

Keywords: Multi-agent systems, Cooperative and distributed learning
Abstract: This paper proposes a novel zeroth-order method for inverse optimal control, utilizing the AMSBound algorithm combined with a two-point gradient estimator. The method aims to learn the parameters in both the system model and the cost function by minimizing a loss function derived from demonstrated trajectories. Unlike existing frameworks that compute the gradient of the loss function by differentiating Pontryagin’s Maximum Principle, the proposed method employs the zeroth-order estimation to approximate the gradient of the loss function. Theoretical guarantees for convergence of the algorithm are provided, and experimental results demonstrate the effectiveness of the proposed method.

Keywords: Multi-agent systems, Nonlinear dynamics over networks
Abstract: In this paper, we develop a neural network-based method to study opinion behaviors under a covariance-based dissensus algorithm. Driven by this dissensus algorithm, the opinions are updated based on relative interactions and gradually converge to dissensus on the sphere. This proposed neural network-based method samples data and trains a neural network to ensure the Lyapunov conditions, which significantly simplifies the Lyapunov function design for stability analysis. The regions of attraction for different dissensus equilibria can also be estimated under opinion dynamics on a unit sphere by training a neural network to approximate the solution of Zubov's equation. Simulations demonstrate the performance of the proposed method.

Keywords: Sensor networks and distributed processing, Multi-agent systems, Robotics and multi-agent systems
Abstract: This paper introduces the obstacle avoidance adaptive coverage control for the mobile sensor network (MSN). In addition to obstacle avoidance, parameter convergence of the unknown spatial sensory function is guaranteed under the mild excitation condition termed collective initial excitation (C-IE). The proportional-integral (PI) type adaptation enables the multi-agents to achieve convergence and consensus under a less stringent excitation condition than collective persistence of excitation (C-PE). The proposed coverage algorithm is evaluated through simulation.

Keywords: Smart cities and power systems, Distributed optimization
Abstract: We consider the problem of temperature regulation and optimal energy sharing in district heating systems, without having a knowledge of the demand, while also maintaining a good transient performance. We show that convex optimization problems can be formulated for jointly achieving optimal temperature regulation and energy sharing. Furthermore, we show that LQR formulations can converge to this optimal operating point, without having a knowledge of the demand, while minimising also prescribed transient performance metrics. The effectiveness of our approach is also validated through comprehensive simulations.

Keywords: Smart cities and power systems, Nonlinear dynamics over networks, Decentralized control and large-scale systems
Abstract: This paper shows that a necessary and sufficient condition for the small-signal stability of an inverter-integrated power system can be expressed in terms of semidefinite matrix inequalities determined only by the synchronous reactance of the components, the susceptance matrix of the transmission network, and the stationary values of the power flow distribution. To derive the stability condition, we consider a class of grid-forming inverters corresponding to a singular perturbation of the synchronous generator. The resulting matrix inequality condition, which has twice as many dimensions as the number of buses and is independent of the dynamics of the connected components, is expressed in terms of each component compensating in a decentralized manner for the loss of frequency synchronization caused by the reactive power consumption in the transmission network. A simple numerical example using a 3-bus power system model shows that a grid-forming inverter load improves power system synchronization, while a grid-following inverter load disrupts it.

Keywords: Cyber security in networked control systems, Multi-agent systems, Cooperative and distributed learning
Abstract: Resilience, understood as the ability of a network to carry out its goals under adversarial attacks and unexpected failures, is critical for autonomy. Despite important advances in the design of distributed coordination and decision-making algorithms, multi-agent networks have proven fragile to targeted attacks. Novel theories and tools are therefore needed to guarantee resiliency of these systems, being the development of notions and techniques that characterize network resilience critical. However, obtaining such characterizations is difficult as resilience and performance are a complex function of the network’s and adversary’s capabilities, knowledge, resources, and the network interconnection structure. At the same time, we also need novel design methodologies that can protect multi-agent networks and adaptively manage their interconnection over time to achieve performance guarantees. In this talk, we present our recent progress in these directions including algorithmic methods for metric computation, data-driven algorithms for multi-agent learning, and defense strategies.

Keywords: Algebraic and geometric methods for network control, Applications of consensus and gossip algorithms, Biological networks and epidemics dynamics
Abstract: This talk explores distributed and decentralized optimization, where computational agents collaborate across networks to solve problems of the form min r(x)+1/n ∑ (f_i(x)+h_i(x)). We introduce a novel approach that reformulates these optimization problems to embed network structures in the constraints, enabling the direct application of mature algorithms. This problem-centric shift — moving away from traditional algorithm-centric methods like modifying gradient descent for distribution — simplifies convergence analysis, clarifies algorithmic relationships, and extends solutions to dynamic networks. To take advantages of dynamic networks, we examine finite-time consensus, where agents achieve agreement in a fixed number of steps using efficient communication protocols, and show how it enhances efficiency within our framework. This joint work with: Xin Jiang, Edward Nguyen, and Bicheng Yin.

Keywords: Applications of consensus and gossip algorithms
Abstract: In this paper, we propose a hierarchical architecture for distributed time scale generation for a cluster of atomic clocks. The objective is to ensure synchronized and accurate timekeeping in both normal and emergency situations. Concretely, in the first layer, a consensus-type control strategy is employed to achieve distributed synchronization. The resultant synchronized state (SS), which represents the generated time scale in this specific application, is controlled in the second layer. In normal situations, SS is gradually steered towards external references. While in emergency situations, floating control is applied to suppress the divergence rate of the unobservable SS. It aims to stabilize the difference between the current SS and a target, which has the least Allan variance for a longer time period. The proposed architecture provides a clear and coherent solution for distributed atomic timekeeping.

Keywords: Applications of consensus and gossip algorithms, Multi-agent systems, Distributed constrained control and MPC
Abstract: This work aims to achieve output consensus of constrained heterogeneous multi-agent systems under a switching network, where outputs are periodic and characterized by a reference system. Compared to constant references, the periodic references are harder to track and to achieve constrained consensus. In this paper, constraint admissible set for the agent-reference system is first analyzed to determine the set of admissible references. This is followed by a consensus protocol involving a projection so that consensus is achieved among the set of admissible references for each system. An additional feature of the proposed approach is the use of an artificial reference such that each agent tracks the given admissible reference without violating constraints even when reference is far from the current state. Finally, the efficiency of the proposed algorithm is demonstrated by numerical examples.

Keywords: Applications of consensus and gossip algorithms, Sensor networks and distributed processing
Abstract: The convergence of the gossip process has been extensively studied; however, algorithms that generate a set of stochastic matrices, the infinite product of which converges to a rank-one matrix determined by a given weight vector, have been less explored. In this work, we propose an algorithm for constructing (local) stochastic matrices based on a given gossip network topology and a set of weights for averaging across different consensus clusters, ensuring that the gossip process converges to a finite limit set.

Keywords: Biological networks and epidemics dynamics, Multi-agent systems, Social and economic networks
Abstract: This paper proposes an SIR (susceptible-infected-recovered) epidemic model influenced by opinion dynamics and explore the effectiveness of opinion-based intervention strategies in controlling epidemic peaks. For a well-mixed population, we derive a theoretically optimal intervention strategy that minimizes the cumulative cost of opinion interventions while maintaining the infection proportion below a specified threshold. This strategy is shown to consist of three phases: no intervention, moderate intervention, and stopping intervention. Extending our results to networked populations, we simulate the optimal intervention strategy on a dual-layer network comprising a physical transmission network (based on inter-prefecture migration data in Japan) and a small-world social network. The numerical simulations demonstrate the effectiveness of the strategy in large-scale network systems.

Keywords: Control over wireless, Control under communication constraints
Abstract: Control loops closed over wireless links greatly benefit from accurate estimates of the communication channel condition. To this end, the finite-state Markov channel model allows for reliable channel state estimation. Our paper, recently accepted for publication in Automatica, has developed a Markov jump linear system representation for wireless networked control with persistent channel state observation, stochastic message losses, and generalized packet dropout compensation. With this model, we solved the finite- and infinite-horizon linear quadratic regulation problems and introduced an easy-to-test stability condition for any given infinite-horizon control law. We also thoroughly analyzed the impact of a scalar general dropout compensation factor on the stability and closed-loop performance of a rotary inverted pendulum controlled remotely through a wireless link. Finally, we validated the results numerically via extensive Monte Carlo simulations, showing the benefits of the proposed control strategy.

Keywords: Control over wireless, Control under communication constraints
Abstract: We propose a cooperative Model Predictive Control (MPC) framework designed for implementation over the edge-cloud continuum. The framework integrates a cloud MPC harnesses the high computational power of the cloud to generate optimal control inputs using a high-fidelity nonlinear model, explicitly accounting for packet losses and communication delays. Moreover the proposed framework includes an edge MPC, which leverages a simplified linear dynamic model to provide rapid feedback while operating under limited computational resources. By intelligently reconciling control signals from both edge and cloud computing devices, this novel edge-cloud MPC framework strikes a balance between computational efficiency and real-time responsiveness, optimizing overall system performance.

Keywords: Control over wireless, Smart cities and power systems, Cooperative and distributed learning
Abstract: Machine learning has proven highly effective in addressing constrained optimization problems by approximating the mapping from hyperparameters to solutions. However, standard supervised learning methods often fall short due to the presence of multiple (sub-)optimal solutions. To address this challenge, we propose a diversity-aware augmented learning framework. Our approach transforms the one-to-many input-solution mapping into a function through the augmentation of the input space with initial points, thereby respecting the diversity of high-quality solutions. The proposed framework enhances the quality and diversity of optimal solution estimation, as evidenced by two case studies.

Keywords: Control under communication constraints, Algebraic and geometric methods for network control, Nonlinear dynamics over networks
Abstract: In this paper, we consider a remote control problem based on data-driven control with an emphasis on communication constraints. Specifically, we propose a direct data-driven stabilization method with quantization in input and state data for unknown discrete-time linear systems. Moreover, the controller is designed taking account of the effects of quantization in the feedback data. Logarithmic type quantization is employed, and we show the inherent trade-off in the quantization coarseness for data-driven design and feedback control. We illustrate the effectiveness of the method through numerical simulations.

Keywords: Cooperative and distributed learning, Distributed optimization, Multi-agent systems
Abstract: In this work, we study federated learning, a distributed learning framework where data remains decentralized across multiple clients, and a shared model is trained collaboratively via a central server. We take a control-theoretic approach, focusing on a specific class of residual neural networks by modeling them as dynamical systems. Building on this perspective, we propose FedProject, an algorithm designed to mitigate statistical heterogeneity caused by non-identically distributed data. Our method introduces a projection operator on the proximal term of FedProx and employs a two-step update to balance local learning and client drift from the global model. We evaluate FedProject against FedProx and FedAvg on both IID and heterogeneous datasets, demonstrating improved convergence and robustness to hyperparameter selection for this class of neural networks.

Keywords: Cyber security in networked control systems
Abstract: This paper is extracted from a recently accepted journal manuscript in IEEE Transactions on Automatic Control (doi: 10.1109/TAC.2023.3321248). This paper designs the cooperative attack strategy for switched systems when the sensor-to-controller and controller-to-actuator channels are compromised by DoS attack and deception attack. The proposed attack strategy is expected to be as stealthy as possible, regardless of whether system mode is eavesdropped or not.

Keywords: Distributed estimation, Multi-agent systems, Cyber security in networked control systems
Abstract: We study resilient distributed parameter estimation in multi-agent systems where some agents may malfunction. The objective is for each nonfaulty agent to locally estimate its parameter while it may interact with adversaries. To this end, we develop an algorithm using multi-hop relaying to achieve the goal in multi-agent networks with directed topologies. With multi-hop relays, agents can access more information of remote agents even though they communicate with only direct neighbors. We characterize a necessary and sufficient graph condition for our algorithm to succeed, which is denoted by the notion of robust following graphs. We prove that our condition with multi-hop relays is more relaxed than the one with one-hop case, and hence, our approach can tolerate more adversaries in the same network when multi-hop relays are applied. Lastly, numerical examples verify the efficacy of our algorithm.

Keywords: Game theory and network games, Graph-theoretic methods for control, Multi-agent systems
Abstract: This paper provides the sensitivity analysis of Linear Quadratic Gaussian graphon mean-field games (LQG-GMFG), with a particular focus on how perturbations in initial conditions at different network locations affect system behavior. We quantify the impact of localized perturbations through a L^2-perturbation metric via graphon spectral decompositions and establish explicit solutions for the perturbation analysis that reveal how network topology influences perturbation propagation patterns. Our theoretical results reveal fundamental connections among network topology, system dynamics, and sensitivity patterns, providing insights for robust network design and control strategies.

Keywords: Graph-theoretic methods for control, Algebraic and geometric methods for network control
Abstract: We establish necessary and sufficient conditions for polynomial systems with a unique visible monomial observable to be (strongly) super-linearized using monomial observables. Our analysis reveals that, while a super-linearization may include additional observables, the observables that are essential for achieving the super-linearization must all have the same degree as the visible observable.

Keywords: Infinite-dimensional systems, Decentralized control and large-scale systems, Algebraic and geometric methods for network control
Abstract: Estimation of stochastic systems on very large networks is intractable computationally. Graphon theory provides limit objects for infinite sequences of graphs by mapping adjacency matrices to the unit square, enabling the modelling of dynamical systems on arbitrarily large graphs via functional analytic methods. In previous work (Dunyak and Caines, 2022, 2023, 2024), Q-noise was used to extend stochastic systems on large graphs to stochastic systems in Hilbert spaces on graphons; in this paper the linear system state estimation problem on large networks and their graphon limits are analysed, and a Separation Principle of control and estimation is introduced. Convergence of finite network linear system state estimates, together with the corresponding Kalman filter systems to their graph limit counterparts, is established. A computational example of this convergence is illustrated on a standard graphon example.

Keywords: Multi-agent systems, Nonlinear dynamics over networks, Robotics and multi-agent systems
Abstract: We introduce an opinion dynamics-based approach to Multi-Robot Task Allocation (MRTA), modeling task assignment as a convergence process of continuously evolving opinions. Our method integrates special task requirements, ensuring that each task receives a minimum number of mobile robots while dynamically adapting to system changes. We also introduce a two-time-scale framework where opinion dynamics operate on a faster time scale to facilitate rapid task allocation, while mobile robots move on a slower time scale, maintaining a stable communication graph during opinion convergence. Additionally, we develop a specialized velocity controller that utilizes location and task assignment information from neighboring robots, promoting the reassignment of robots to under-selected tasks. We validate our strategy through simulations, demonstrating its effectiveness across diverse robot and task settings. Our findings highlight the potential of opinion dynamics in enhancing the scalability, efficiency, and reliability of distributed MRTA systems for mobile robot swarms.

Keywords: Nonlinear dynamics over networks, Applications of consensus and gossip algorithms, Emergent behavior
Abstract: We propose a novel bias-based consensus framework for nonlinear opinion dynamics. Due to the observable and malleable nature of bias in human-robot interactions, we utilize it as a control parameter to achieve consensus. First, we analyze the Lyapunov–Schmidt reduced system near equilibrium under small bias assumptions. Through constrained cusp bifurcation, we show that increasing individual biases beyond identified thresholds—and relative biases beyond saddle-node limit points ensures consensus with a unique stable equilibrium. For large biases, we conduct a global phase-plane analysis. By establishing strong monotonicity and applying the Poincaré–Bendixson theorem, we eliminate the possibility of limit cycles and guarantee consensus with a unique stable attractor as equilibrium. Finally, along with numerical simulations for the two-agent, two-option case, we show that the proposed bias control approach extends seamlessly to decentralized multi-agent opinion consensus.

Keywords: Robotics and multi-agent systems, Distributed constrained control and MPC
Abstract: In this paper, we propose a lightweight online embedded solver for input-constrained model predictive control problems on low-power, resource-constrained microcontrollers. Our method combines gradient projection and conjugate gradient techniques to quickly identify the optimal working set within an active-set strategy. Both simulations and real-world tests confirm that the proposed solver yields optimized inputs with reduced overshoot and enhanced smoothness, demonstrating its effectiveness on resource-limited hardware.

Keywords: Emergent behavior, Robotics and multi-agent systems
Abstract: Navigating mobile robots in a turbulent flow field presents significant challenges due to unpredictable odorant plume dispersion and intermittent environmental cues. This paper presents a reinforcement learning (RL)-based approach for robotic source-seeking in such environments, inspired by fruit flies' navigation behaviors. A Deep Q-Network (DQN) model is trained using experimentally recorded trajectories of fruit flies to develop an adaptive search strategy. The robot learns to make navigation decisions based on limited sensory feedback, leveraging stochastic environmental cues to improve its movement toward the source. The RL-based approach demonstrates its ability to generalize across different trajectories, achieving higher accumulated rewards than biological trajectories. Simulation results demonstrate the model’s robustness and adaptability, highlighting the potential of RL for bio-inspired navigation in mobile robotics and environmental monitoring.

Keywords: Decentralized control and large-scale systems
Abstract: We tackle the problem of stabilizing unknown interconnected systems and propose data-driven decentralized control algorithms. We first derive a data-driven condition to synthesize a local controller that ensures the dissipativity of the local subsystem. Then, we propose a data-driven decentralized stability condition for the global system based on the dissipativity of each local system. Since both conditions take the form of linear matrix inequalities, this yields a unified pipeline, resulting in a decentralized data-driven control algorithm. We also consider systems interconnected through diffusive coupling and propose a more decentralized and scalable control algorithm. We provide numerical examples in the context of microgrids to validate the effectiveness and the scalability of the algorithms.

Keywords: Decentralized control and large-scale systems, Smart cities and power systems, Multi-agent systems
Abstract: We propose a novel analysis framework for the small-signal stability of grid-forming converters in future power systems. Our approach leverages dynamic loop-shifting techniques to compensate for the lack of passivity in the network dynamics and establishes decentralized parametric stability certificates, depending on the local device-level controls and incorporating the effects of the interconnecting line dynamics. Through carefully designed tuning rules, we are able to ensure plug-and-play operation without centralized coordination. Unlike prior works, our approach accommodates coupled frequency and voltage dynamics, incorporates line dynamics, and does not rely on specific network configurations or operating points, offering a general and scalable solution for the integration of power-electronics based units into future power systems.

Keywords: Multi-agent systems, Applications of consensus and gossip algorithms, Cyber security in networked control systems
Abstract: In this talk, we provide an overview on the recent advances in the research of multi-agent systems operating in hostile environments. We will will focus on the influence of misbehaving agents in a network capable to inject false data in their transmissions and how to mitigate such attacks by the approach based on the socalled mean subsequence reduced algorithms and their variants. Agents equipped with such algorithms will ignore their neighbors taking extreme state values. We will see that characterizations on the properties necessary for network topologies have been established, and moreover that network resiliency can be enhanced when more communication and computational resources are available. We will further discuss extensions of such algorithms to problems of averaging, leader-follower consensus, parameter estimation, and clock synchronization in wireless sensor networks.