This paper addresses bipartite output consensus of heterogeneous multi-agent systems (MASs) communicating over signed directed networks. A distributed adaptive observation mechanism is proposed to jointly estimate the leader trajectory and dynamics from signed relative information available to individual agents, without using network-wide spectral quantities. The coordination objective is then embedded into a cooperative output regulation framework. Since the estimated leader matrix renders the regulator equations time varying, an online adaptive solver with auxiliary states is formulated to generate their solutions. These auxiliary states improve transient convergence while keeping the computation tractable as the matrix dimension increases. The observer and solver outputs are incorporated into two distributed controller implementations suited to different local information conditions. Stability analysis establishes the asymptotic convergence of the regulated outputs to the desired bipartite trajectories. Examples concerning multi-robot positioning and microgrid active-power tracking corroborate the analytical results and demonstrate the convergence improvement achieved by the proposed solver.
This paper studies the distributed optimal resource allocation problem for high-order systems on unbalanced digraphs, where agents' decisions are constrained by equality-coupled network resource constraints. The objective is to control agents to autonomously minimize the sum of their local cost functions while satisfying the network resource constraints. Existing distributed resource allocation algorithms are often limited by the complex coexistence of network imbalance and high-order agent dynamics. To address this challenge, a novel distributed algorithm is developed based on eigenvector estimators and adaptive control methods, eliminating the need for prior knowledge of the global information about network connectivity. Rigorous analysis is given to ensure that the high-order systems with the proposed algorithm can achieve optimal resource allocation. Finally, simulations are conducted to validate the effectiveness and advantages of the proposed approach.
In this paper, we mainly investigate three important problems for target controllability of complex networks, with a particular focus on practical input constraints where driver nodes can only be selected from a predefined subset rather than freely across the entire network. Under such constraints, networks are often not structurally controllable. This necessitates minimizing the number of driver nodes while maximizing the number of controllable target nodes, leading to the constrained target controllability problem. Building upon this, we consider two additional related problems: constrained multilevel target controllability, where the target node set is divided into multiple hierarchical subsets, and the objective is to maximize the number of controllable target nodes of each subset in a levelwise order, and constrained minimum-cost target controllability, which focuses on minimizing the total cost of control paths and cycles. To address these problems, we propose a novel unified framework called “minimum-cost flow based target path cover” (MCTP). Simulation results on Erdős–Rényi, scale-free, and real-world networks demonstrate the effectiveness and scalability of the MCTP framework. Additionally, we introduce two new measures to evaluate the robustness of target controllability with MCTP. The versatility of the MCTP framework suggests that it offers a powerful and generalizable tool for analyzing target controllability problems across a wide range of networks.
This paper proposes a distributed adaptive consensus control design technique with guaranteed parameter convergence for a class of nonlinear multi-agent systems. First, adaptive distributed observers with prescribed performance guarantees are developed to estimate the leader's states, thereby decoupling the consensus control design problem into distinct observer and controller layers. This layered structure significantly simplifies the design of follower controllers, each of which depends solely on its own local states and the observer estimates, eliminating the need for neighbor state information. Subsequently, in the controller layer, both historical data from individual agents and real-time inter-agent shared data are integrated into cooperative adaptive update laws, enabling dual-dimensional (temporal-spatial) data-driven cooperative concurrent learning. An integral-type barrier Lyapunov function that explicitly incorporates adaptive parameter estimation errors is introduced, facilitating a unified stability analysis for both the observer and controller layers. Under a newly proposed cooperative data full-rank condition, which is less restrictive than conventional cooperative persistent excitation and finite excitation conditions, we rigorously prove that the adaptive parameter estimates of all agents converge exponentially to their common ideal values, achieving cooperative identification of unknown parameters in the follower dynamics. Numerical simulations are provided to demonstrate the effectiveness of the proposed technique.
This paper deals with the problem of establishing global asymptotical stability (GAS) of a rational system under denial-of-service (DoS) attacks constrained in terms of average frequency and duration. Our approach is based on the search for a pair of polynomial Lyapunov functions via convex optimization techniques, where the conservatism decreases with the degree of the Lyapunov functions. Firstly, a sufficient condition is proposed for GAS, which amounts to establishing if a cost function is negative for at least one triplet of parameters that satisfies a system of bilinear matrix inequalities (BMIs). This condition is also necessary under some assumptions in the case of homogeneous rational systems of relative degree one. Secondly, it is shown how candidate parameters certifying GAS can be obtained by solving three quasi-convex optimization problems in the form of generalized eigenvalue problems (GEVPs), which may solve the problem or may be used to construct a bounding region for the sought parameters. Thirdly, a sufficient condition for determining parameters certifying GAS is proposed based on the solution of two convex optimization problems in the form of semidefinite programs (SDPs) constructed with the introduction of polynomial parameter-dependent Lyapunov functions, which is also necessary under some assumptions in the case of homogeneous rational systems of relative degree one.
