Communication Graph Research Articles

Cooperative Decisions of a Multi-Agent System for the Target-Pursuit Problem in Manned–Unmanned Environment

With the development of intelligent technology, multi-agent systems have been widely applied in military and civilian fields. Compared to a single platform, multi-agent systems can complete more dangerous, difficult, and heavy tasks. However, due to the limited autonomy of unmanned platforms and the regulatory needs of personnel, multi-agent systems cooperating with manned platforms to perform tasks have been more widely promoted at this stage of development. This paper addresses a differential game method for cooperative decision-making of a multi-agent system cooperating with the manned platform for the target-pursuit problem. The manned platform pursues the target according to a certain trajectory, and its state can be obtained by the multi-agent system. Firstly, for the case that the target moves with a fixed trajectory, the target-pursuit problem in a manned–unmanned environment is viewed in the form of game based on a communication graph among agents. Secondly, strategies of all agents are proposed while maintaining their group cohesion. A set of coupled differential equations is solved to implement strategy calculation. Compared to purely unmanned systems, the strategies combine the advantages of the manned platform and add a reference item, which can achieve team cohesion relatively quickly. Furthermore, a brief analysis is made on the scenarios where the target is in another case or adopts other strategies. Finally, comparative simulations have verified the effectiveness and synergy of the strategy.

Distributed optimization of multi-integrator agent systems with mixed neighbor interactions

In this paper, we consider a constrained distributed optimization problem for multi-integrator agent systems with mixed neighbor interactions, where the neighbors of the agents can be classified as cooperative or noncooperative by introducing a cooperation-competition network into the modeling of the communication graph. The agent’s goal is to optimize the total objective function of the cooperators. Since agents can only communicate with their neighbors, they are unable to obtain complete information about cooperators. The distributed optimization algorithm is proposed to classify cooperative subnetworks by introducing additional auxiliary variables, and the outputs of all agents exponentially converge to the optimal solution of the optimization problem under the condition of satisfying the global equality constraint of the subnetwork. The Lyapunov stability theory is used to analyze the algorithm’s convergence. Simulation results show the effectiveness of the proposed algorithm in two examples of economic dispatch in smart grids and optimal area coverage in multi-sensor networks.

Bearing‐only containment control of multi‐agent systems

AbstractBearing‐only containment control means to achieve containment using only the bearing information, rather than the complete relative position information including both the bearing and the distance. Compared with the most existing relative position based containment control algorithms, the bearing‐only containment control algorithm is nonlinear due to the nonlinearity of the bearing, which renders the stability analysis challenging. In this article, the bearing‐only containment control problem of multi‐agent systems with dynamically changing topologies is investigated for two cases, where the leaders are stationary or dynamic. First, when the leaders are stationary, it is shown that the nonlinearity of the bearing leads to the unknown nonlinear time‐varying parameters in the system. Based on the boundedness of the nonlinear parameters and the connectivity of the switching directed sensing graphs, it is proven that all the followers converge to the stationary convex hull formed by the leaders asymptotically. Second, when the leaders are dynamic, a distributed estimator is proposed for each follower to keep a local estimate of the velocity of the leaders. Based on the optimality of the projection and the shift invariance of the normal cone, it is proven that all the followers converge to the dynamic convex hull formed by the leaders asymptotically under the conditions on the switching directed sensing and communication graphs. It is shown from the two cases that the distance information is not necessary for containment control, and containment can be achieved as long as the bearing information is provided. Numerical examples are presented to illustrate the theoretical results.

A robust distributed observer design for Lipschitz nonlinear systems with time-varying switching topology

This paper deals with state estimation for a class of Lipschitz nonlinear systems under a time-varying disconnected communication network. A distributed observer consists of some local observers that are connected to each other through a communication network. We consider a situation where a communication network does not remain connected all the time, and the network may be caused by intermittent communication link failure. Moreover, each local observer has access to a local measurement, which may be insufficient to ensure the system’s observability, but the collection of all measurements in the network ensures observability. In this condition, the purpose is to design a distributed observer where the estimated state vectors of all local observers converge to the state vector of the system asymptotically, while local observers exchange estimated state vectors through a communication network and use their local measurements. According to theoretical analysis, a nonlinear and a robust nonlinear distributed observer exist when in addition to the union of all communication topologies being strongly connected during a time interval, the component of each communication graph is also strongly connected during each subinterval. The existence conditions of the distributed observers are derived in terms of a set of linear matrix inequalities (LMIs). Finally, the effectiveness of the presented method is numerically verified using some simulation examples.

Distributed k-Winners-Take-All Network: An Optimization Perspective.

In this article, we proposed an equivalent formulation of the k-winners-take-all (k-WTA) problem as a constrained optimization problem by including the Laplacian matrix of the undirected connected communication graph to adapt to the distributed computing scenario, where an additional auxiliary variable is introduced. To solve the optimization problem in a distributed fashion, we design projection neural networks by using the convex optimization theory, leading to the emergence of a distributed k-WTA network. Our theoretical analysis shows that the proposed distributed k-WTA network has a globally asymptotically stable equilibrium that is identical to the optimal solution to the optimization problem, that is, the correct k-WTA solution. The effectiveness and advantages, including the extendability to constrained k-WTA problems, of the proposed k-WTA network are demonstrated via simulations.

Event-driven fully distributed optimal coordinated control for Euler–Lagrange multi-agent systems with connectivity preservation

This paper studies the distributed optimal coordinated control problem for Euler–Lagrange multi-agent systems with connectivity preservation. The aim is to force agents to achieve the optimal solution minimizing the sum of the local objective functions while guaranteeing the connectivity of the communication graph. For practical purposes, the gradient vector of the local objective function is allowed to use only at the real-time generalized position instead of at the auxiliary system state. To make the control parameters independent of the global information and guarantee the fully distributed manner of controller, the adaptive control is introduced to update the coupling weights of the relative states among neighbors. Moreover, to reduce the resource for control updates, the event-driven communication is employed for the updates of both the relative states and the gradient of the connectivity-preserving potential function. Based on the Lyapunov analysis framework, it is proved that agents can converge to the optimal solution with connectivity preservation and Zeno behavior is excluded for the two event-triggering conditions. Finally, the effectiveness of the proposed method is verified by a numerical simulation example.

Fully distributed bipartite output formation tracking for heterogeneous singular multiagent systems on signed digraphs

AbstractThis article investigates the bipartite output formation tracking problem of heterogeneous singular multiagent systems (MASs) over a signed digraph on which two types of cases without and with input saturation are considered, respectively. First, a novel formation tracking protocol is presented for heterogeneous singular MASs in the absence of input saturation constraints. In the proposed protocol, a distributed observer with adaptive updating mechanism is constructed for each follower, which does not involve global information of communication graph. Based on the observer, the designed controller addresses bipartite output formation tracking of heterogeneous singular MASs in a fully distributed fashion. Then, by the output regulation theory and low‐gain feedback technique, a low‐gain formation tracking protocol is proposed to achieve semi‐global bipartite output formation tracking for heterogeneous singular MASs subject to input saturation constraints. Finally, two simulation examples are given to illustrate the feasibility of theoretical results.

Multi-UAV Roundup Inspired by Hierarchical Cognition Consistency Learning Based on an Interaction Mechanism

This paper is concerned with the problem of multi-UAV roundup inspired by hierarchical cognition consistency learning based on an interaction mechanism. First, a dynamic communication model is constructed to address the interactions among multiple agents. This model includes a simplification of the communication graph relationships and a quantification of information efficiency. Then, a hierarchical cognition consistency learning method is proposed to improve the efficiency and success rate of roundup. At the same time, an opponent graph reasoning network is proposed to address the prediction of targets. Compared with existing multi-agent reinforcement learning (MARL) methods, the method developed in this paper possesses the distinctive feature that target assignment and target prediction are carried out simultaneously. Finally, to verify the effectiveness of the proposed method, we present extensive experiments conducted in the scenario of multi-target roundup. The experimental results show that the proposed architecture outperforms the conventional approach with respect to the roundup success rate and verify the validity of the proposed model.

Generalized Semistability and Stochastic Semistability for Switched Nonlinear Systems Using Fixed Point Theory

Abstract This paper generalizes our previous results on semistability and stochastic semistability for switched nonlinear systems published in the Proceedings of 2021 Modeling, Estimation and Control Conference. The paper also provides the results on semistability in mean square for switched nonlinear discrete-time systems. The theoretical result involves generalized sufficient conditions for (stochastic) semistability and semistability in mean square of discrete-time nonlinear dynamical systems under time-varying or random (arbitrary) switching by means of Fixed Point Theory. An advantage of these results is to overcome fundamental challenges arising from using existing methods such as Lyapunov and LaSalle methods. As an application of the theoretical results presented, a constrained distributed consensus problem over random multi-agent networks is considered for which a generalized asynchronous and totally asynchronous iterative algorithm is derived. The algorithm is able to converge even if the weighted matrix of the graph is periodic and irreducible under synchronous protocol. Finally, a numerical example is given in which there is a distribution dependency among communication graphs to demonstrate the results.

Output Consensus of Heterogeneous Linear Multiagent Systems With Directed Graphs via Adaptive Dynamic Event-Triggered Mechanism

This article investigates the output consensus problem of heterogeneous linear multiagent systems under directed communication graphs. A novel adaptive dynamic event-triggered mechanism is proposed, intending to further save system resources consumed in communication between agents and controller update of agents themselves, and remove the assumption that the global information associated with the communication topology should be known in advance for the design of control parameters. Unlike the existing related adaptive event-triggered algorithms, the proposed algorithm could theoretically guarantee the existence of a strictly positive constant on the interevent time intervals for both communication between agents and controller update. Furthermore, it is shown by the simulation that the addition of a dynamic variable has the potential to further optimize the control cost when compared to the addition of exponential function signal or L₁ signal usually adopted in existing adaptive event-triggered mechanisms. Then, the obtained results are extended from the strongly connected graph to the directed communication graph only containing a spanning tree. Finally, numerical simulation results are conducted to demonstrate the effectiveness of the proposed mechanism.

Distributed consensus control for voltage tracking and current distribution in DC microgrid

DC microgrids are extensively researched because of advancements in power electronics, penetration of DC loads, and the growing demand for DC power. The DC microgrid is also broached as a multi-agent system (MAS) because of its distributed nature. In this paper, a distributed consensus controller for DC microgrids is proposed that has basically two main objects; the first object is to track the desired reference voltage for the distributed generation units while maintaining the voltage stability at the point of common coupling (PCC), whereas the second object is to distribute current proportionally among distributed loads. Furthermore, a linear distributed model is constructed in which the dynamics of the net current for the distributed generation units are established using the directed graph of power lines. Moreover, local information from neighboring agents is utilized to construct a distributed linear controller, and the consensus control problem is solved for voltage tracking and current distribution of the DC microgrid. Furthermore, the control gains for the proposed controller are optimized using closed-loop stability analysis. Finally, simulation results are illustrated for both open-loop and closed-loop systems, and the effects of the directed power line graph and the undirected communication graph are analyzed. The effects of parameter uncertainties on the performance of the proposed controller are investigated through simulations. The results show that the distributed controller efficiently solves the consensus control problem for the MAS.

Fault Injection Based Interventional Causal Learning for Distributed Applications

We apply the machinery of interventional causal learning with programmable interventions to the domain of applications management. Modern applications are modularized into interdependent components or services (e.g. microservices) for ease of development and management. The communication graph among such components is a function of application code and is not always known to the platform provider. In our solution we learn this unknown communication graph solely using application logs observed during the execution of the application by using fault injections in a staging environment. Specifically, we have developed an active (or interventional) causal learning algorithm that uses the observations obtained during fault injections to learn a model of error propagation in the communication among the components. The “power of intervention” additionally allows us to address the presence of confounders in unobserved user interactions. We demonstrate the effectiveness of our solution in learning the communication graph of well-known microservice application benchmarks. We also show the efficacy of the solution on a downstream task of fault localization in which the learned graph indeed helps to localize faults at runtime in a production environment (in which the location of the fault is unknown). Additionally, we briefly discuss the implementation and deployment status of a fault injection framework which incorporates the developed technology.

Privacy-preserving Decentralized Federated Learning over Time-varying Communication Graph

Establishing how a set of learners can provide privacy-preserving federated learning in a fully decentralized (peer-to-peer, no coordinator) manner is an open problem. We propose the first privacy-preserving consensus-based algorithm for the distributed learners to achieve decentralized global model aggregation in an environment of high mobility, where participating learners and the communication graph between them may vary during the learning process. In particular, whenever the communication graph changes, the Metropolis-Hastings method [ 69 ] is applied to update the weighted adjacency matrix based on the current communication topology. In addition, the Shamir’s secret sharing (SSS) scheme [ 61 ] is integrated to facilitate privacy in reaching consensus of the global model. The article establishes the correctness and privacy properties of the proposed algorithm. The computational efficiency is evaluated by a simulation built on a federated learning framework with a real-world dataset.

Upgrading edges in the Graphical TSP

In this paper we present a generalization of the Graphical Traveling Salesman Problem (GTSP). Given a communication graph in which not all direct connections are necessarily possible, the Graphical Traveling Salesman Problem consists of finding the shortest tour that visits each node at least once. In this work, we assume the availability of a budget that allows to upgrade, i.e. reduce their traversal cost, some of the current connections and we propose the problem of designing the minimum cost tour using this budget. We propose and study a formulation for the problem, verifying that the polyhedron associated with the set of feasible solutions of a relaxed version of the problem is a full-dimensional polytope. We present families of valid inequalities that reinforce the model and pre-processing techniques to reduce the number of variables of the formulation. To solve the problem, we propose a branch-and-cut algorithm that uses the introduced valid inequalities, as well as a heuristic to obtain good upper bounds and a tailor-made branching strategy. Comprehensive computational experiments on a new set of benchmark instances are presented to assess the performance of this exact method.

Bipartite consensus of multi-agent systems with matched uncertainty via fully distributed edge-based event-triggered mechanism

The leader-following bipartite consensus of multi-agent systems (MASs) with matched uncertainty is investigated by using the fully distributed asynchronous edge-based event-triggered mechanism. Firstly, event-triggered mechanisms are constructed for each edge and the leader to decrease the consumption of system resources. The state feedback and output feedback control protocols are proposed, which do not depend on the global values of the communication graph. Secondly, sufficient conditions for the bipartite consensus of MASs are obtained by Lyapunov stability theory. Thirdly, the feasibility of the proposed event-triggered mechanisms is further verified by exclusion of Zeno phenomenon. Finally, the effectiveness of control protocol is illustrated by simulation.

Distributed event‐triggered secondary control for microgrids applicable to directed communication graph

AbstractRecently, distributed cooperative control has been popularly used in the secondary control of microgrids due to its reliability and flexibility. However, traditional distributed control schemes based on fixed‐cycle communication usually result in a waste of communication resources. Thus, this paper proposes an event‐triggered distributed control strategy that can significantly reduce the requirements of communication. In addition, compared with existing schemes, the proposed algorithm can achieve the secondary control goal in a directed graph, which can be applied to more scenarios, especially for weaker communication conditions. The stability of the strategy is investigated with the Lyapunov method. The simulation results verify the validity and effectiveness of the proposed scheme.

Communication area estimation for decentralized control of nanosatellites swarm

The problem of relative drift elimination between the satellites in the swarm is considered in the paper. The proposed decentralized control takes into account a communication constraint such as limited size of communication area. Only the satellites within the communication area can be identified by relative motion determination system. The control aim is to eliminate the mean relative drift between all the satellites inside the communication area. The purpose of the work is to study the performance of the proposed decentralized control algorithm. It is shown that the system matrix of differential equations for the vector of relative drifts is related to the Laplacian matrix of the communication graph. In the case of the connected swarm, all but one eigenvalues of the system are negative, and the remaining one is equal to zero. It means that all the relative drifts converge to the same value under the proposed control. The speed of convergence is defined by the minimum absolute eigenvalue that depends on the graph topology. The initial drift and the convergence speed make it possible to estimate the communication distance that provides the connectivity of the graph. Considering normally distributed errors of the initial velocity after the launch, it is possible to estimate the distance between any two satellites after the convergence. It allows us to estimate the communication distance that ensures the relative drift elimination between all the satellites in the swarm. The obtained estimations are validated using Monte Carlo simulations. In numerical simulations the swarm of 3U CubeSats in low-Earth orbit is considered. The decentralized control is implemented by differential aerodynamic drag via the change of cross-sectional area using onboard reaction wheels.

A novel consensus-oriented distributed optimization scheme with convergence analysis for economic dispatch over directed communication graphs

A novel consensus-oriented distributed optimization scheme with convergence analysis for economic dispatch over directed communication graphs

Distributed Disturbance Observer-Based Containment Control of Multi-Agent Systems via an Event-Triggered Approach

This paper studies the containment control problem of linear multi-agent systems (MASs) subject to external disturbances, where the communication graph is a directed graph with the followers being undirected connections. In order to save communication costs and energy consumption, a distributed disturbance observer-based event-triggered controller is employed based on the relative outputs of neighboring followers. Compared with conventional controllers, our observer-based controller utilizes the relative outputs of neighboring followers at the same triggered instant. Furthermore, it is shown that Zeno behavior can be avoided. Finally, the validity of our proposed methodology is demonstrated by a simulation example.

Multiagent Low-Dimensional Linear Bandits

We study a multi-agent stochastic linear bandit with side information, parameterized by an unknown vector <formula><tex>$\theta^* \in \mathbb{R}^d$</tex></formula>. The side information consists of a finite collection of low-dimensional subspaces, one of which contains <formula><tex>$\theta^*$</tex></formula>. In our setting, agents can collaborate to reduce regret by sending recommendations across a communication graph connecting them. We present a novel decentralized algorithm, where agents communicate subspace indices with each other and each agent plays a projected variant of LinUCB on the corresponding (low-dimensional) subspace. By distributing the search for the optimal subspace across users and learning of the unknown vector by each agent in the corresponding low-dimensional subspace, we show that the per-agent finite-time regret is much smaller than the case when agents do not communicate. We finally complement these results through simulations.