Issue In Multi-agent Systems Research Articles

Performance Analysis of Fault-Tolerant Multiagent Coordination Mechanisms

Performance evaluation of multi-agent systems (MAS) embraces several challenges due to uncertain operational environments, such as software/hardware failures and unfaithful communications that facilitate the spread of deceptive messages. One way to smooth the impact of reliability and security potential issues in MAS is to enforce different coordination mechanisms among agents (i.e., coordinating multiple agents that need to perform a sequence of actions to maximize a system-level reward), and evaluate their efficiency. In this paper, we propose a tool-based approach, namely COORDINATE, that simulates and compares the performance characteristics of different coordination mechanisms for MAS while considering fault-tolerant and corrupted agents. A Smart Hospital is adopted as illustrative example to show the need of performance-based analysis results pointing out which coordination mechanism is more efficient. Experimental results indicate that the inter-arrival time of tasks to be accomplished, the failure probability of agents, and the ratio of faithful to malicious agents contribute to determining the efficiency of different coordination mechanisms. When varying these parameters in the considered scenarios, the system latency can be reduced up to 4.2× by selecting the optimal coordination mechanism. This way, software developers are informed on the system peculiarities that trigger the switching among such coordination mechanisms for a performance-based optimal solution.

An Effective Negotiation Strategy for Quantitative and Qualitative Issues in Multi-Agent Systems

Automated negotiation is an efficient approach for interaction in multi-agent systems in which agents exchange offers and counteroffers to conclude an agreement. This paper addresses the problem of offer formulation during the interaction between buyer and seller software agents for the purpose of reaching an agreement over quantitative and qualitative issues at once. In order to improve the outcome of the negotiation process, a hybrid negotiation method is presented and verified. Offer formulation is based on fuzzy similarity and preference-based methods. The preference-based mechanism is used for quantitative issues, while the fuzzy similarity technique is used for qualitative issues. The preference-based mechanism takes into account the preferences of the opponent when generating offers; the agent makes greater concessions on the issues which the opponent prefers more. The fuzzy-similarity method formulates an offer that considers offering a deal that is more similar to the one received by the opponent during the last round of negotiation. The experiments consists of two parts. The first part compares the hybrid strategy with the basic one. The findings reveal that the hybrid strategy is better in all performance measures, namely, utility rate, agreement rate, and Nash product rate. The second part of the experimental work compares four mechanisms of offer generating mechanisms: basic, preference-based, fuzzy similarity, and hybrid. The results show that the hybrid negotiation strategy performs equal or better that other negotiation strategies. More details can be found in the paper.

Hybrid-order Network Consensus for Distributed Multi-agent Systems

As an important field of Distributed artificial intelligence (DAI), multi-agent systems (MASs) have attracted the attention of extensive research scholars. Consensus as the most important issue in MAS, much progress has been made in studying the consensus control of MAS, but there are some problems remained largely unaddressed which cause the MAS to lose some useful network structure information. First, multi-agent consensus protocol usually proceeds over the low-order structure by only considering the direct edges between agents, but ignores the higher-order structure of the whole topology network. Second, the existing work assumes all the edges in a topology network have the same weight without exploring the potential diversity of the connections. In this way, multi-agent systems fail to enforce consensus, resulting in fragmentation into multiple clusters. To address the above issues, this paper proposes a Motif-aware Weighted Multi-agent System (MWMS) method for consensus control. We focus more on triangle motif in the network, but it can be extended to other kinds of motifs as well. First, a novel weighted network is used which is the combination of the edge-based lower-order structure and the motif-based higher-order structure, i.e., hybrid-order structure. Subsequently, by simultaneously considering the quantity and the quality of the connections in the network, a novel consensus framework for MAS is designed to update agents. Then, two baseline consensus algorithms are used in MWMS. In our experiments, we use ten topologies of different shapes, densities and ranges to comprehensively analyze the performance of our proposed algorithms. The simulation results show that the hybrid higher-order network can effectively enhance the consensus of the multi-agent system in different network topologies.

Recursive Sliding-Mode Dynamic Surface Containment Control With Nonlinear Gains for Uncertain Nonlinear Multiagent Systems

Containment control is one of critical issues in multiagent systems (MASs). A recursive sliding-mode dynamic surface control (DSC) strategy with a nonlinear gain is addressed for the containment problem of uncertain MASs in strict-feedback form. Two factors, a recursive dynamic sliding-mode surface and a nonlinear gain function, are introduced to solve the contradiction between control accuracy and dynamic performance in the traditional DSC method. The function approximation technique, by radial basis function neural networks, is employed to compensate unknown dynamics. On the basis of the backstepping and graph theory, a full-state feedback containment control strategy is presented. Subsequently, we propose an output feedback containment control strategy via a state observer for individual followers. From the Lyapunov stability theorem, we construct an improvement Lyapunov function for the MAS, and it is proven that all followers converge to a convex hull formed by leaders for both full-state feedback and output feedback cases. Simulation examples are presented for verification of the effectiveness of the strategy.

Two approximation algorithms for probabilistic coalition structure generation with quality bound

How to form effective coalitions is an important issue in multi-agent systems. Coalition Structure Generation ({{mathsf {CSG}}}) is a fundamental problem whose formalization can encompass various applications related to multi-agent cooperation. {{mathsf {CSG}}} involves partitioning a set of agents into coalitions such that the social surplus (i.e., the sum of the values of all coalitions) is maximized. In traditional {mathsf {CSG}}, we are guaranteed that all coalitions will be successfully established, that is, the attendance rate of each agent for joining any coalition is assumed to be 1.0. Having the real world in mind, however, it is natural to consider the uncertainty of agents’ availabilities, e.g., an agent might be available only two or three days a week because of his/her own schedule. Probabilistic Coalition Structure Generation ({{mathsf {PCSG}}}) is an extension of {mathsf {CSG}} where the attendance type of each agent is considered. The aim of this problem is to find the optimal coalition structure which maximizes the sum of the expected values of all coalitions. In {mathsf {PCSG}}, since finding the optimal coalition structure easily becomes intractable, it is important to consider approximation algorithms, i.e., to consider a trade-off between the quality of the returned solution and tractability. In this paper, a formal framework for {mathsf {PCSG}} is introduced. Approximation algorithms for {mathsf {PCSG}} called Bounded Approximation Algorithm based on Attendance Types ({{mathsf {BAAAT}}}) and Involved {mathsf {BAAAT}} ({{mathsf {IBAAAT}}}) are then presented. We prove a priori bounds on the quality of the solution returned by {mathsf {BAAAT}} and {mathsf {IBAAAT}} with respect to the optimum and perform experimental evaluations on a number of benchmarks.

Task Allocation for Multi-Agent Systems Based on Distributed Many-Objective Evolutionary Algorithm and Greedy Algorithm

Task allocation is a key issue in multi-agent systems, and finding the optimal strategy for task allocation has been proved to be an NP-hard problem. Existing task allocation methods for multi-agent systems mainly adopt distributed full search strategies or local search strategies. The former requires a lot of computation and communication costs, while the latter cannot ensure the diversity and quality of solutions. Therefore, in this paper, we combine a distributed many-objective evolutionary algorithm called D-NSGA3 with a greedy algorithm to search the task allocation solutions, and we comprehensively consider the constraints related to space, time, energy consumption and agent function in multi-agent systems. Specifically, D-NSGA3 is used to optimize multiple objectives simultaneously so as to ensure the search capability and the diversity of solutions. Alternate search between D-NSGA3 and the greedy algorithm is conducted to enhance the local optimizing ability. Experimental results show that the proposed method can effectively solve large-scale task allocation problems (e.g., the number of agents is not less than 250, and that of tasks is not less than 1000). Compared with the existing work called MSEA, the proposed method could achieve better and more diverse solutions.

ACoPla: a Multiagent Simulator to Study Individual Strategies in Dynamic Situations

One important issue in multi-agent systems is how to define agents’ interaction strategies in dynamic open environments. Generally, agents’ behaviors, such as being cooperative/altruistic or competitive/adversarial, are defined a priori by their creators. However, this is a weak premise when considering interaction among anonymous self-interested agents. Whenever agents meet, there is always a decision to be made: what is the best group interaction strategy? We argue that the answer depends on the amount of information required to make a decision and on the deadline proximity for accomplishing the task in hand. In certain situations, it is to the agents’ advantage to exchange information with others, while in other situations there are no incentives for them to spend time doing so. Understanding effective behaviors according to the decision- making scenario is still an open issue in multi-agent systems. In this paper, we present a multi-agent simulator (ACoPla) to understand the correlations between agents’ interaction strategy, decision-making context and successful task accomplishment rate. Additionally, we develop a case study in the domain of site evacuation to exemplify our findings. Through this study, we detect the types of conditions under which cooperation becomes the preferred strategy, as the environment changes.

Semantically Enabled Communication Framework for Autonomous Systems

Multi-agent systems (MAS) have recently gained much fame and have proved to be a potential technology that enables the construction and development of collaborative intelligent software systems. Development of such autonomous intelligent systems require a high level of communication with enriched semantics that will enable them to be integrated with the Web to achieve the ultimate goal of the Semantic Web in the future. The most important issues in MAS are high-level agent interaction, expressive power of agents, and semantics of messages to collaborate and cooperate with each other in Autonomous Decentralized Systems (ADS). This encoding/decoding mechanism will not only increase the reliability of the system but will also exhibit the properties of ADS that are high assurance. The evaluation proves that the implementation of semantically enriched communication improves performance and efficiency as a whole.

A Survey of Self-Organization Mechanisms in Multiagent Systems

This paper surveys the literature over the last decades in the field of self-organizing multiagent systems. Self-organization has been extensively studied and applied in multiagent systems and other fields, e.g., sensor networks and grid systems. Self-organization mechanisms in other fields have been thoroughly surveyed. However, there has not been a survey of self-organization mechanisms developed for use in multiagent systems. In this paper, we provide a survey of existing literature on self-organization mechanisms in multiagent systems. We also highlight the future work on key research issues in multiagent systems. This paper can serve as a guide and a starting point for anyone who will conduct research on self-organization in multiagent systems. Also, this paper complements existing survey studies on self-organization in multiagent systems.

Scalable multi-agent learning algorithms to determine winners in combinatorial double auctions

Many complex decision problems in the real world are very difficult to solve due to computational complexity. The winner determination problem (WDP) in combinatorial double auctions is one example of such complex problems. Although it allows buyers and sellers to submit bids and trade goods conveniently, the WDP in combinatorial double auctions is notoriously difficult to solve from computation point of view. It relies on the development of an effective mechanism to determine the winners. Multi-agent systems (MAS) provide an approach in which several agents attempt, through their interactions, to jointly solve a problem. An important issue in MAS is the design of multi-agent learning algorithms. In this paper, we will study the development of scalable multi-agent learning algorithms for solving the WDP in combinatorial double auctions. Instead of finding the exact solution, we will set up a fictitious market based on MAS architecture and develop multi-agent learning algorithms to reduce the computational complexity in solving the WDP of combinatorial double auctions. In the fictitious market, each buyer, each seller and the mediator is modeled by an agent. The issue is to develop learning algorithms for all the agents in the system to collectively solve the WDP in combinatorial double auctions. In this paper, we adopt a Lagrangian relaxation approach and a subgradient method to develop efficient multi-agent learning algorithms for solving the WDP in combinatorial double auctions. The effectiveness of the proposed multi-agent learning algorithms is also demonstrated by numerical examples.

Containment control of multi-agent systems with time delay

This paper studies the containment control issue of multi-agent systems with time delay. By adopting continuous-time and sampled-data-based control protocols, we establish some necessary and sufficient containment control criteria for multi-agent systems with time delay. For the continuous-time case, our results show that both topology structure and time delay play an important role in the containment control of multi-agent systems, no matter whether the leaders are stationary (first-order system) or dynamic with constant velocity and time-varying position (second-order system). Furthermore, the maximum value of time delay to guarantee the achievement of containment control is given by the Hopf bifurcation theory. Moreover, we extend the results to the sampled-data case. Finally, numerical simulations are given to illustrate our main results.

Magentix2: A privacy-enhancing Agent Platform

Agent Platforms are the software that supports the development and execution of Multi-agent Systems. There are many Agent Platforms developed by the agent community, but they hardly consider privacy. This leads to agent-based applications that invade users’ privacy. Privacy can be threatened by two main information activities: information collection and information processing. Information collection can be prevented using traditional security mechanisms. Information processing can be prevented by minimizing data identifiability, i.e., the degree by which personal information can be directly attributed to a particular individual. However, minimizing data identifiability may directly affect other crucial issues in Multi-agent Systems, such as accountability, trust, and reputation. In this paper, we present the support that the Magentix2 Agent Platform provides for preserving privacy. Specifically, it provides mechanisms to avoid information collection and information processing when they are not desired. Moreover, Magentix2 provides these mechanisms without compromising accountability, trust, and reputation. We also provide in this paper an application built on top of Magentix2 that exploits its support for preserving privacy. Finally, we provide an extensive evaluation of the support that Magentix2 provides for preserving privacy based on that application. We specifically test whether or not privacy loss can be minimized by using the support that Magentix2 provides, whether or not this support introduces a bearable performance overhead, and whether or not existing trust and reputation models can be implemented on top of Magentix2.

μ-SATPLAN: Multi-agent planning as satisfiability

Planning is a fundamental issue in multi-agent systems. In this work we focus on the coordination of multiple agents in two different settings. In the first, agents are able to attain individual goals that are necessary for the achievement of a global common goal. As the agents share the same environment, they need to find a coordinated course of action that avoids harmful (or negative) interactions, and benefit from positive interactions, whenever this is possible. In the second setting some of the agents may need the assistance of other agents to achieve their individual goals. This is the case where some of the actions of the plan of an agent may be executed by another agent who will play the role of the assistant. We formalize these two problems in a more general way than in previous works, and present a coordination algorithm which generates optimal solutions in the case of two agents. In this algorithm, agents use μ-SATPLAN as the underlying planner for generating individual and joint consistent plans. This planner is an extension of the classical SATPLAN planner, that tackles negative and positive interactions and, therefore, multi-agent planning. We also present an algorithm that solves the assistance problem. The underlying algorithm is again μ-SATPLAN, and is used for the generation of individual (based on assistance) and joint consistent plans. Finally experimental results on multi-agent versions of problems taken from International Planning Competitions demonstrate the effectiveness of μ-SATPLAN and the coordination algorithm.

A survey of security issue in multi-agent systems

Multi-agent systems have attracted the attention of researchers because of agents' automatic, pro-active, and dynamic problem solving behaviors. Consequently, there has been a rapid development in agent technology which has enabled us to provide or receive useful and convenient services in a variety of areas such as banking, transportation, e-business, and healthcare. In many of these services, it is, however, necessary that security is guaranteed. Unless we guarantee the security services based on agent-based systems, these services will face significant deployment problems. In this paper, we survey existing work related to security in multi-agent systems, especially focused on access control and trust/reputation, and then present our analyses. We also present existing problems and discuss future research challenges.

Incorporating trust into the BDI architecture

Trust is becoming a major issue in Multi-Agent Systems (MAS), in which agents need to delegate some tasks to others. Trust can be defined as the degree of confidence in an agent's prediction of another agent's future behaviour. In the context of the BDI model, trust can be defined as the beliefs that one agent has of another's beliefs, desires, intentions and capabilities. This paper first presents our own version of BDI interpreter. Then, the above vision of trust is incorporated into the BDI interpreter in order to help an agent in the decision process before delegating tasks to other agents.

Approximate Strong Equilibrium in Job Scheduling Games

A Nash Equilibrium (NE) is a strategy profile resilient to unilateral deviations, and is predominantly used in the analysis of multiagent systems. A downside of NE is that it is not necessarily stable against deviations by coalitions. Yet, as we show in this paper, in some cases, NE does exhibit stability against coalitional deviations, in that the benefits from a joint deviation are bounded. In this sense, NE approximates strong equilibrium. Coalition formation is a key issue in multiagent systems. We provide a framework for quantifying the stability and the performance of various assignment policies and solution concepts in the face of coalitional deviations. Within this framework we evaluate a given configuration according to three measures: (i) IR_min: the maximal number alpha, such that there exists a coalition in which the minimal improvement ratio among the coalition members is alpha, (ii) IR_max: the maximal number alpha, such that there exists a coalition in which the maximal improvement ratio among the coalition members is alpha, and (iii) DR_max: the maximal possible damage ratio of an agent outside the coalition. We analyze these measures in job scheduling games on identical machines. In particular, we provide upper and lower bounds for the above three measures for both NE and the well-known assignment rule Longest Processing Time (LPT). Our results indicate that LPT performs better than a general NE. However, LPT is not the best possible approximation. In particular, we present a polynomial time approximation scheme (PTAS) for the makespan minimization problem which provides a schedule with IR_min of 1+epsilon for any given epsilon. With respect to computational complexity, we show that given an NE on m >= 3 identical machines or m >= 2 unrelated machines, it is NP-hard to determine whether a given coalition can deviate such that every member decreases its cost.

Computational logic-based agents

Computational logic provides a rigorous framework for designing, specifying, implementing and verifying autonomous agents that interact, communicate with one another and coordinate their activities in multi-agent systems and societies. In particular, computational logic can aid formalizing reasoning with agents’ mental attitudes such as beliefs, desires, intentions and goals, and agents’ social attitudes such as protocols, social commitments, obligations, norms and rules of behaviour. The use of computational logic aims at facilitating the development of agents that reason and act rationally while at the same time being able to verify the behaviour of these agents against their specification. In the last decade, computational logic-based agents have gained popularity and many contributions have been made in this field both from a theoretical and a practical viewpoint. Computational logic has been used to provide powerful techniques and tools to address a number of issues in multi-agent systems, ranging from agent communication, to agent programming languages, to argumentation, to decision making in a single and multi-agent context, to applying model checking for verifying agent systems. The further development and integration of these techniques and tools shows promise for playing a fundamental role in developing powerful and verifiable agent-based systems. This special issue is about using computational logic to deal with some key issues in multi-agent systems. The purpose is to report on recent developments in this area and discuss current challenges, limits, and future research opportunities. The problems discussed range from agent programming languages through a BDI agent-based programming language, to agent communication through a multi-issue bilateral negotiation, to logic programming-based agents realising a distributed abductive inference mechanism, to mental attitudes representation through a theory of intention representation, and automated, nondeterministic planning for extended reachability goals in agents.

Controlling non-normative behaviors by anticipation for autonomous agents

One of the crucial issues in multi-agent systems is the coordination of agents' actions. A simple way to coordinate the agents is to define norms in the system. Most of the research related to norms and multi-agent systems focuses on designing normative agents to ensure a social order inside the system. To achieve this goal, these studies make the assumption that non-normative behavior must be controlled using sanction mechanisms. This paper discusses the relevance of this assumption in certain specific contexts, and presents an alternative for controlling non-normative behavior, thus ensuring the social order. To illustrate our proposal, we introduce a concrete application involving traffic simulation at a road junction. Based on the statistical traffic results, we show how our approach helps to improve the realism of the simulation.

An improved trust model in agent-mediated e-commerce

Trust is a fundamental issue in multiagent systems, especially in agent-mediated e-commence. The trust model plays an important role in determining who and how to interact in open and dynamic environments. To this end, many trust models have been developed. Based on the confidence reputation model proposed by other researchers, an improved trust model is discussed. The original model was enhanced in two aspects: (1) when evaluating the trustworthiness of target agents, the reliability of the witness itself is taken into account and further aggregated by Dempster-Shafer evidence theory and (2) the ontological property of trust is considered, which implies that trust can be calculated more precisely. A case study is provided to show the effectiveness of the improved model.

A swarm based approach to adapt the structural dimension of agents' organizations

One of the well studied issues in multi-agent systems is the standard action-selection problem where a goal task can be performed in different ways, by different agents. Also the sequence of these actions can influence the goal achievement or its quality. This class of problems has been tackled under different approaches. At the high-level coordination one, the specification of the organizational issues is crucial. However, in dynamic environments, agents must be able to adapt to the changing organizational goals, available resources, their relationships to the presence of another agents, and so on. This problem is a key one in multi-agent systems and relates to models of learning and adaptation, such as those observed among social insects. The present paper tackles the process of generating, adapting, and changing multi-agent organization dynamically at system runtime, using a swarm inspired approach. This approach is used here mainly for task allocation with low need of pre-planning and specification, and no need of explicit coordination. The results of our approach and another quantitative one are compared here and it is shown that in dynamic domains, the agents adapt to changes in the organization, just as social insects do.