An Optimal Control-Based Distributed Reinforcement Learning Framework for A Class of Non-Convex Objective Functionals of the Multi-Agent Network

This paper studies a novel distributed optimization problem that aims to minimize the sum of the non-convex objective functionals of the multi-agent network under privacy protection, which means that the local objective of each agent is unknown to others. The above problem involves complexity simultaneously in the time and space aspects. Yet existing works about distributed optimization mainly consider privacy protection in the space aspect where the decision variable is a vector with finite dimensions. In contrast, when the time aspect is considered in this paper, the decision variable is a continuous function concerning time. Hence, the minimization of the overall functional belongs to the calculus of variations. Traditional works usually aim to seek the optimal decision function. Due to privacy protection and non-convexity, the Euler-Lagrange equation of the proposed problem is a complicated partial differential equation.Hence, we seek the optimal decision derivative function rather than the decision function. This manner can be regarded as seeking the control input for an optimal control problem, for which we propose a centralized reinforcement learning(RL) framework. In the space aspect, we further present a distributed reinforcement learning framework to deal with the impact of privacy protection. Finally, rigorous theoretical analysis and simulation validate the effectiveness of our framework.

Event-Triggered Differentially Private Average Consensus for Multi-agent Network

This paper investigates the differentially private problem of the average consensus for a class of discrete-time multi-agent network systems(MANSs). Based on the MANSs,a new distributed differentially private consensus algorithm(DPCA) is developed. To avoid continuous communication between neighboring agents, a kind of intermittent communication strategy depending on an event-triggered function is established in our DPCA. Based on our algorithm, we carry out the detailed analysis including its convergence, its accuracy, its privacy and the trade-off between the accuracy and the privacy level, respectively. It is found that our algorithm preserves the privacy of initial states of all agents in the whole process of consensus computation. The trade-off motivates us to find the best achievable accuracy of our algorithm under the free parameters and the fixed privacy level. Finally, numerical experiment results testify the validity of our theoretical analysis.

Target Tracking and Obstacle Avoidance for Multi-agent Networks with Input Constraints

In this paper, the problems of target tracking and obstacle avoidance for multi-agent networks with input constraints are investigated. When there is a moving obstacle, the control objectives are to make the agents track a moving target and to avoid collisions among agents. First, without considering the input constraints, a novel distributed controller can be obtained based on the potential function. Second, at each sampling time, the control algorithm is optimized. Furthermore, to solve the problem that agents cannot effectively avoid the obstacles in dynamic environment where the obstacles are moving, a new velocity repulsive potential is designed. One advantage of the designed control algorithm is that each agent only requires local knowledge of its neighboring agents. Finally, simulation results are provided to verify the effectiveness of the proposed approach.

Multi-Agent Network Intrusion Active Defense Model Based on Immune Theory

Inspired by the immune theory and multi-agent systems, an immune multi-agent active defense model for network intrusion is established. The concept of immune agent is introduced, and its running mechanism is established. The method, which uses antibody concentration to quantitatively describe the degree of intrusion danger, is presented. This model implements the multi-layer and distributed active defense mechanism for network intrusion. The experiment results show that this model is a good solution to the network security defense.

Leader-following consensus protocols for formation control of multi-agent network

Two protocols are presented,which can make agents reach consensus while achieving and preserving the desired formation in fixed topology with and without communication timedelay for multi-agent network.First,the protocol without considering the communication time-delay is presented,and by using Lyapunov stability theory,the sufficient condition of stability for this multi-agent system is presented.Further,considering the communication time-delay,the effectiveness of the protocol based on Lyapunov-Krasovskii function is demonstrated.The main contribution of the proposed protocols is that,as well as the velocity consensus is considered,the formation control is concerned for multi-agent systems described as the second-order equations.Finally,numerical examples are presented to illustrate the effectiveness of the proposed protocols.

Leader-following formation control of multi-agent networks based on distributed observers

To investigate the leader-following formation control, in this paper we present the design problem of control protocols and distributed observers under which the agents can achieve and maintain the desired formation from any initial states, while the velocity converges to that of the virtual leader whose velocity cannot be measured by agents in real time. The two cases of switching topologies without communication delay and fixed topology with time-varying communication delay are both considered for multi-agent networks. By using the Lyapunov stability theory, the issue of stability is analysed for multi-agent systems with switching topologies. Then, by considering the time-varying communication delay, the sufficient condition is proposed for the multi-agent systems with fixed topology. Finally, two numerical examples are given to illustrate the effectiveness of the proposed leader-following formation control protocols.

Pinning consensus analysis of multi-agent networks with arbitrary topology

In this paper the pinning consensus of multi-agent networks with arbitrary topology is investigated. Based on the properties of M-matrix, some criteria of pinning consensus are established for the continuous multi-agent network and the results show that the pinning consensus of the dynamical system depends on the smallest real part of the eigenvalue of the matrix which is composed of the Laplacian matrix of the multi-agent network and the pinning control gains. Meanwhile, the relevant work for the discrete-time system is studied and the corresponding criterion is also obtained. Particularly, the fundamental problem of pinning consensus, that is, what kind of node should be pinned, is investigated and the positive answers to this question are presented. Finally, the correctness of our theoretical findings is demonstrated by some numerical simulated examples.

Collective surrounding control in multi-agent networks

We study the collective encirclement control problem of a multi-agent system in this paper, where the agents move collectively to encircle the multiple targets. An algorithm is proposed and a sufficient condition is derived to realize the collective encirclement motion. Some simulation results are provided to show the effectiveness of the obtained theoretical results.

Rigidity based formation tracking for multi-agent networks

This paper considers the formation tracking problem under a rigidity framework, where the target formation is specified as a minimally and infinitesimally rigid formation and the desired velocity of the group is available to only a subset of the agents. The following two cases are considered： the desired velocity is constant, and the desired velocity is timevarying. In the first case, a distributed linear estimator is constructed for each agent to estimate the desired velocity. The velocity estimation and a formation acquisition term are employed to design the control inputs for the agents, where the rigidity matrix plays a central role. In the second case, a distributed non-smooth estimator is constructed to estimate the time-varying velocity, which is shown to converge in a finite time. Theoretical analysis shows that the formation tracking problem can be solved under the proposed control algorithms and estimators. Simulation results are also provided to show the validity of the derived results.

Finite-time Prescribed Performance Time-Varying Formation Control for Second-Order Multi-Agent Systems With Non-Strict Feedback Based on a Neural Network Observer

This paper studies the problem of time-varying formation control with finite-time prescribed performance for nonstrict feedback second-order multi-agent systems with unmeasured states and unknown nonlinearities.To eliminate nonlinearities,neural networks are applied to approximate the inherent dynamics of the system.In addition,due to the limitations of the actual working conditions,each follower agent can only obtain the locally measurable partial state information of the leader agent.To address this problem,a neural network state observer based on the leader state information is designed.Then,a finite-time prescribed performance adaptive output feedback control strategy is proposed by restricting the sliding mode surface to a prescribed region,which ensures that the closed-loop system has practical finite-time stability and that formation errors of the multi-agent systems converge to the prescribed performance bound in finite time.Finally,a numerical simulation is provided to demonstrate the practicality and effectiveness of the developed algorithm.

CoopAI-Route: DRL Empowered Multi-Agent Cooperative System for Efficient QoS-Aware Routing for Network Slicing in Multi-Domain SDN

The emergence of beyond 5G networks has the potential for seamless and intelligent connectivity on a global scale.Network slicing is crucial in delivering services for different,demanding vertical applications in this context.Next-generation applications have time-sensitive requirements and depend on the most efficient routing path to ensure packets reach their intended destinations.However,the existing IP(Internet Protocol)over a multi-domain network faces challenges in enforcing network slicing due to minimal collaboration and information sharing among network operators.Conventional inter-domain routing methods,like Border Gateway Protocol(BGP),cannot make routing decisions based on performance,which frequently results in traffic flowing across congested paths that are never optimal.To address these issues,we propose CoopAI-Route,a multi-agent cooperative deep reinforcement learning(DRL)system utilizing hierarchical software-defined networks(SDN).This framework enforces network slicing in multi-domain networks and cooperative communication with various administrators to find performance-based routes in intra-and inter-domain.CoopAI-Route employs the Distributed Global Topology(DGT)algorithm to define inter-domain Quality of Service(QoS)paths.CoopAI-Route uses a DRL agent with a message-passing multi-agent Twin-Delayed Deep Deterministic Policy Gradient method to ensure optimal end-to-end routes adapted to the specific requirements of network slicing applications.Our evaluation demonstrates CoopAI-Route’s commendable performance in scalability,link failure handling,and adaptability to evolving topologies compared to state-of-the-art methods.

Safety-Constrained Multi-Agent Reinforcement Learning for Power Quality Control in Distributed Renewable Energy Networks

This paper examines the difficulties of managing distributed power systems,notably due to the increasing use of renewable energy sources,and focuses on voltage control challenges exacerbated by their variable nature in modern power grids.To tackle the unique challenges of voltage control in distributed renewable energy networks,researchers are increasingly turning towards multi-agent reinforcement learning(MARL).However,MARL raises safety concerns due to the unpredictability in agent actions during their exploration phase.This unpredictability can lead to unsafe control measures.To mitigate these safety concerns in MARL-based voltage control,our study introduces a novel approach:Safety-ConstrainedMulti-Agent Reinforcement Learning(SC-MARL).This approach incorporates a specialized safety constraint module specifically designed for voltage control within the MARL framework.This module ensures that the MARL agents carry out voltage control actions safely.The experiments demonstrate that,in the 33-buses,141-buses,and 322-buses power systems,employing SC-MARL for voltage control resulted in a reduction of the Voltage Out of Control Rate(%V.out)from0.43,0.24,and 2.95 to 0,0.01,and 0.03,respectively.Additionally,the Reactive Power Loss(Q loss)decreased from 0.095,0.547,and 0.017 to 0.062,0.452,and 0.016 in the corresponding systems.

Multilayer Satellite Network Collaborative Mobile Edge Caching:A GCN-Based Multi-Agent Approach

With the explosive growth of highdefinition video streaming data,a substantial increase in network traffic has ensued.The emergency of mobile edge caching(MEC)can not only alleviate the burden on core network,but also significantly improve user experience.Integrating with the MEC and satellite networks,the network is empowered popular content ubiquitously and seamlessly.Addressing the research gap between multilayer satellite networks and MEC,we study the caching placement problem in this paper.Initially,we introduce a three-layer distributed network caching management architecture designed for efficient and flexible handling of large-scale networks.Considering the constraint on satellite capacity and content propagation delay,the cache placement problem is then formulated and transformed into a markov decision process(MDP),where the content coded caching mechanism is utilized to promote the efficiency of content delivery.Furthermore,a new generic metric,content delivery cost,is proposed to elaborate the performance of caching decision in large-scale networks.Then,we introduce a graph convolutional network(GCN)-based multi-agent advantage actor-critic(A2C)algorithm to optimize the caching decision.Finally,extensive simulations are conducted to evaluate the proposed algorithm in terms of content delivery cost and transferability.

基于Multi-Agent的无人机集群体系自主作战系统设计

针对无人集群自主作战体系设计中的关键问题,提出基于Multi-Agent的无人集群自主作战系统设计方法。建立无人集群各节点的Agent模型及其推演规则;对于仿真系统模块化和通用化的需求,设计系统互操作式接口和无人集群自主作战的交互关系;开展无人集群系统仿真推演验证。仿真结果表明,所提设计方案不仅能够有效开展并完成自主作战网络生成-集群演化-效能评估的全过程动态演示验证,而且能够通过重复随机试验进一步评估无人集群的协同作战效能,最后总结了集群协同作战的策略和经验。

基于Multi-Agent的水电站变压器故障诊断系统

为了精准、快速完成水电站变压器的故障诊断,设计基于Multi-Agent的水电站变压器故障诊断系统。变压器状态监控agent将检测到的变压器故障信息发送给系统管理agent,系统管理agent通过通信agent将变压器故障信息发送给变压器故障诊断agent,变压器故障诊断agent利用小波变换方法提取变压器故障特征,并将其作为IFOA-SVM模型输入,完成变压器故障分类后,获取变压器故障诊断结果,该结果通过通信agent显示给用户。实验表明,该系统可有效诊断变压器故障诊断,诊断成功率受系统故障信息丢失率的影响较小,诊断耗时、耗能小,并具有较高故障诊断成功率。

UAV-Assisted Dynamic Avatar Task Migration for Vehicular Metaverse Services: A Multi-Agent Deep Reinforcement Learning Approach

Avatars, as promising digital representations and service assistants of users in Metaverses, can enable drivers and passengers to immerse themselves in 3D virtual services and spaces of UAV-assisted vehicular Metaverses. However, avatar tasks include a multitude of human-to-avatar and avatar-to-avatar interactive applications, e.g., augmented reality navigation,which consumes intensive computing resources. It is inefficient and impractical for vehicles to process avatar tasks locally. Fortunately, migrating avatar tasks to the nearest roadside units(RSU)or unmanned aerial vehicles(UAV) for execution is a promising solution to decrease computation overhead and reduce task processing latency, while the high mobility of vehicles brings challenges for vehicles to independently perform avatar migration decisions depending on current and future vehicle status. To address these challenges, in this paper, we propose a novel avatar task migration system based on multi-agent deep reinforcement learning(MADRL) to execute immersive vehicular avatar tasks dynamically. Specifically, we first formulate the problem of avatar task migration from vehicles to RSUs/UAVs as a partially observable Markov decision process that can be solved by MADRL algorithms. We then design the multi-agent proximal policy optimization(MAPPO) approach as the MADRL algorithm for the avatar task migration problem. To overcome slow convergence resulting from the curse of dimensionality and non-stationary issues caused by shared parameters in MAPPO, we further propose a transformer-based MAPPO approach via sequential decision-making models for the efficient representation of relationships among agents. Finally, to motivate terrestrial or non-terrestrial edge servers(e.g., RSUs or UAVs) to share computation resources and ensure traceability of the sharing records, we apply smart contracts and blockchain technologies to achieve secure sharing management. Numerical results demonstrate that the proposed approach outperforms the MAPPO approach by around 2% and effectively reduces approximately 20% of the latency of avatar task execution in UAV-assisted vehicular Metaverses.

Service Function Chain Deployment Algorithm Based on Multi-Agent Deep Reinforcement Learning

Aiming at the rapid growth of network services,which leads to the problems of long service request processing time and high deployment cost in the deployment of network function virtualization service function chain(SFC)under 5G networks,this paper proposes a multi-agent deep deterministic policy gradient optimization algorithm for SFC deployment(MADDPG-SD).Initially,an optimization model is devised to enhance the request acceptance rate,minimizing the latency and deploying the cost SFC is constructed for the network resource-constrained case.Subsequently,we model the dynamic problem as a Markov decision process(MDP),facilitating adaptation to the evolving states of network resources.Finally,by allocating SFCs to different agents and adopting a collaborative deployment strategy,each agent aims to maximize the request acceptance rate or minimize latency and costs.These agents learn strategies from historical data of virtual network functions in SFCs to guide server node selection,and achieve approximately optimal SFC deployment strategies through a cooperative framework of centralized training and distributed execution.Experimental simulation results indicate that the proposed method,while simultaneously meeting performance requirements and resource capacity constraints,has effectively increased the acceptance rate of requests compared to the comparative algorithms,reducing the end-to-end latency by 4.942%and the deployment cost by 8.045%.

Discovering Latent Variables for the Tasks With Confounders in Multi-Agent Reinforcement Learning

Efficient exploration in complex coordination tasks has been considered a challenging problem in multi-agent reinforcement learning(MARL). It is significantly more difficult for those tasks with latent variables that agents cannot directly observe. However, most of the existing latent variable discovery methods lack a clear representation of latent variables and an effective evaluation of the influence of latent variables on the agent. In this paper, we propose a new MARL algorithm based on the soft actor-critic method for complex continuous control tasks with confounders. It is called the multi-agent soft actor-critic with latent variable(MASAC-LV) algorithm, which uses variational inference theory to infer the compact latent variables representation space from a large amount of offline experience.Besides, we derive the counterfactual policy whose input has no latent variables and quantify the difference between the actual policy and the counterfactual policy via a distance function. This quantified difference is considered an intrinsic motivation that gives additional rewards based on how much the latent variable affects each agent. The proposed algorithm is evaluated on two collaboration tasks with confounders, and the experimental results demonstrate the effectiveness of MASAC-LV compared to other baseline algorithms.

A survey on multi-agent reinforcement learning and its application

Multi-agent reinforcement learning(MARL)has been a rapidly evolving field.This paper presents a comprehensive survey of MARL and its applications.We trace the historical evolution of MARL,highlight its progress,and discuss related survey works.Then,we review the existing works addressing inherent challenges and those focusing on diverse applications.Some representative stochastic games,MARL means,spatial forms of MARL,and task classification are revisited.We then conduct an in-depth exploration of a variety of challenges encountered in MARL applications.We also address critical operational aspects,such as hyperparameter tuning and computational complexity,which are pivotal in practical implementations of MARL.Afterward,we make a thorough overview of the applications of MARL to intelligent machines and devices,chemical engineering,biotechnology,healthcare,and societal issues,which highlights the extensive potential and relevance of MARL within both current and future technological contexts.Our survey also encompasses a detailed examination of benchmark environments used in MARL research,which are instrumental in evaluating MARL algorithms and demonstrate the adaptability of MARL to diverse application scenarios.In the end,we give our prospect for MARL and discuss their related techniques and potential future applications.

Bipartite consensus problems of Lurie multi-agent systems over signed graphs: A contraction approach

This paper examines the bipartite consensus problems for the nonlinear multi-agent systems in Lurie dynamics form with cooperative and competitive communication between different agents. Based on the contraction theory, some new conditions for the nonlinear Lurie multi-agent systems reaching bipartite leaderless consensus and bipartite tracking consensus are presented. Compared with the traditional methods, this approach degrades the dimensions of the conditions, eliminates some restrictions of the system matrix, and extends the range of the nonlinear function. Finally, two numerical examples are provided to illustrate the efficiency of our results.