Interactions In Multi-agent Systems Research Articles

Dynamic Tangled Derivative Logic of Metric Spaces

Dynamical systems are abstract models of interaction between space and time. They are often used in fields such as physics and engineering to understand complex processes, but due to their general nature, they have found applications for studying computational processes, interaction in multi-agent systems, machine learning algorithms and other computer science related phenomena. In the vast majority of applications, a dynamical system consists of the action of a continuous `transition function' on a metric space. In this work, we consider decidable formal systems for reasoning about such structures. Spatial logics can be traced back to the 1940's, but our work follows a more dynamic turn that these logics have taken due to two recent developments: the study of the topological mu-calculus, and the the integration of linear temporal logic with logics based on the Cantor derivative. In this paper, we combine dynamic topological logics based on the Cantor derivative and the `next point in time' operators with an expressively complete fixed point operator to produce a combination of the topological mu-calculus with linear temporal logic. We show that the resulting logics are decidable and have a natural axiomatisation. Moreover, we prove that these logics are complete for interpretations on the Cantor space, the rational numbers, and subspaces thereof.

Network-based kinetic models: Emergence of a statistical description of the graph topology

Abstract In this paper, we propose a novel approach that employs kinetic equations to describe the collective dynamics emerging from graph-mediated pairwise interactions in multi-agent systems. We formally show that for large graphs and specific classes of interactions a statistical description of the graph topology, given in terms of the degree distribution embedded in a Boltzmann-type kinetic equation, is sufficient to capture the collective trends of networked interacting systems. This proves the validity of a commonly accepted heuristic assumption in statistically structured graph models, namely that the so-called connectivity of the agents is the only relevant parameter to be retained in a statistical description of the graph topology. Then, we validate our results by testing them numerically against real social network data.

Investigating Nonlinear Consensus Interaction in Multi-Agent Systems: Exploring Symmetry and Asymmetry

Investigating Nonlinear Consensus Interaction in Multi-Agent Systems: Exploring Symmetry and Asymmetry

Transdisciplinary integration of knowledge in the environment of a virtual stem center

The subject of the article is the role of ontological systems in improving the processes of structuring and analyzing scientific content, especially in the context of STEM education.
 The goal of the work is to research and analyze the application of ontological systems in the context of managing educational materials in STEM education. The use of such approaches is focused on developing effective methods for structuring and presenting educational knowledge in the STEM education system, emphasizing the importance of integrating different scientific disciplines to optimize the educational process. In accordance with the purpose, the following tasks were set: to develop a methodology for creating and implementing ontological systems in STEM education and to develop the architecture of a virtual STEM center that would provide the implementation of the proposed principles.
 The research is based on the following methods: For the development of the architecture of the virtual STEM center, we used UML diagrams. We developed UML diagrams of roles and activities that illustrate the interaction of different users and systems, as well as demonstrate work processes and interactions in multi-agent systems. Special attention is paid to activity diagrams, which reflect the processing of user requests and the interaction of the stemua.science agent with other components of the STEM center.
 The following results were obtained: A modular system architecture of the virtual STEM center was developed and described using UML diagrams, which includes roles such as the STEM center administrator, editor, author, and user, as well as the administrator of the CIT "Polyhedron". The interaction of these roles with the virtual STEM center is described in detail, revealing the mechanisms of their interaction and joint work aimed at creating, filling, and editing content in the transdisciplinary STEM center.
 The process of optimizing work processes in the modular system of the virtual STEM center is also considered. Ways to fill and use the T-STEM center in an ontological form have been identified. The interaction of software entities of the T-STEM center in an ontological form has been analyzed.
 Conclusions: Based on the conducted research, it is concluded that the use of ontological systems in the context of managing educational materials in STEM education is an effective method for structuring and presenting scientific content, promoting the integration of various scientific disciplines, and optimizing the learning process. It is determined that ontological systems are an effective method for structuring and presenting scientific content, facilitating the integration of various scientific disciplines, and optimizing the learning process. The modular architecture of the system is found to facilitate efficient interaction among different roles and automate workflow processes. Integration with a multi-agent system allows for the use of external data sources and ensures interoperability with other systems. For the further development of the system, research is needed to enhance the efficiency of role interactions and workflow automation. Additionally, research on integrating the system with other STEM education systems is necessary.

Commitment-based negotiation semantics for accountability in multi-agent systems

Negotiation is a key form of interaction in multi-agent systems. Negotiation enables agents to come to a mutual agreement about a goal or plan of action. Current negotiation approaches use traditional interaction protocols which do not capture the normative meaning of interactions and often restrict agent autonomy. These traditional negotiation approaches also have difficulty specifying accountability. This paper seeks to address this gap in maintaining autonomy and capturing accountability during negotiation through the use of normative commitments. We propose Nala, a commitment-based negotiation semantics. Nala uses commitments to provide normative meaning to agent interactions. The nature of commitments support in capturing accountability through the violation of created commitments. We illustrate Nala’s usage via a case study using a game scenario where agents participate in negotiation to bring about their goals in a research constrained environment.

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.

Multiscale Generative Models: Improving Performance of a Generative Model Using Feedback from Other Dependent Generative Models

Realistic fine-grained multi-agent simulation of real-world complex systems is crucial for many downstream tasks such as reinforcement learning. Recent work has used generative models (GANs in particular) for providing high-fidelity simulation of real-world systems. However, such generative models are often monolithic and miss out on modeling the interaction in multi-agent systems. In this work, we take a first step towards building multiple interacting generative models (GANs) that reflects the interaction in real world. We build and analyze a hierarchical set-up where a higher-level GAN is conditioned on the output of multiple lower-level GANs. We present a technique of using feedback from the higher-level GAN to improve performance of lower-level GANs. We mathematically characterize the conditions under which our technique is impactful, including understanding the transfer learning nature of our set-up. We present three distinct experiments on synthetic data, time series data, and image domain, revealing the wide applicability of our technique.

Online Health-Conscious Energy Management Strategy for a Hybrid Multi-Stack Fuel Cell Vehicle Based on Game Theory

The use of multiple low-power fuel cells (FCs), instead of a high-power one, in the powertrain of a FC-hybrid electric vehicle (FC-HEV) has recently received considerable attention. This is mainly due to the fact that this configuration can lead to higher efficiency, durability, and reliability. However, the added degrees of freedom require an advanced multi-agent energy management strategy (EMS) for an effective power distribution among power sources. This paper puts forward an EMS based on game theory (GT) for a multi-stack FC-HEV with three FCs and a battery pack. GT is a well-approved method for characterizing the interactions in multi-agent systems. Unlike the other strategies, the proposed EMS is equipped with an online identification system to constantly update the time-varying characteristics of the power sources. The performance of the suggested strategy is investigated through two case studies. Firstly, a comparative study with two other EMSs, dynamic programming (offline), and a competent rule-based strategy (online), is conducted to realize the capability of GT. Secondly, to justify the necessity of online system identification, the degradation effect of each power source on the EMS performance is examined. The carried-out studies show that the total cost (hydrogen consumption and degradation) of the proposed strategy is almost 6% better than the rule-based EMS while keeping a reasonable difference with dynamic programming. Moreover, health unawareness of power sources can increase the hydrogen consumption up to 7% in the studied system.

A Reputation Mechanism for Dynamical Interactions in Multiagent Systems Under Quality of Service Requirements

A novel and original distributed reputation mechanism is here proposed for a class of networked systems characterized by two distinct groups of nodes: user agents and service providers (servers/resources). The proposed methodology is in charge of intercepting those resources that have the highest quality of service (QoS) on the basis of their reputation (trust) shared among agents. Such a reputation may change over the time mainly influenced by changing performance indexes and other events. As a result, it is shown that the trust evolution can be modeled as the output of a positive switching linear system. This model straightforwardly allows one to prove trust boundedness and other relevant steady-state properties. To illustrate the benefits of this approach, a sensors selection problem for state estimation is outlined and addressed within the proposed framework. In particular, it is formally proved that the resulting estimation scheme ensures, over the time, the selection of the sensor with the current best perceived QoS in a finite time.

Generative Attention Networks for Multi-Agent Behavioral Modeling

Understanding and modeling behavior of multi-agent systems is a central step for artificial intelligence. Here we present a deep generative model which captures behavior generating process of multi-agent systems, supports accurate predictions and inference, infers how agents interact in a complex system, as well as identifies agent groups and interaction types. Built upon advances in deep generative models and a novel attention mechanism, our model can learn interactions in highly heterogeneous systems with linear complexity in the number of agents. We apply this model to three multi-agent systems in different domains and evaluate performance on a diverse set of tasks including behavior prediction, interaction analysis and system identification. Experimental results demonstrate its ability to model multi-agent systems, yielding improved performance over competitive baselines. We also show the model can successfully identify agent groups and interaction types in these systems. Our model offers new opportunities to predict complex multi-agent behaviors and takes a step forward in understanding interactions in multi-agent systems.

Dynamics of Generic Linear Agents Over Signed Networks Without Structural Constraints

Signed networks have been widely used to describe cooperative and competitive interactions in multiagent systems (MASs) so far. Most of the existing dynamics results of MASs on signed networks have a certain constraint on the network topology, that is, the network topology is required to have sufficient connectivity for realizing consensus or bipartite consensus. The highlight of this article is to extend the existing dynamics of MASs to a more general signed network, in which there are no any structural constraints on the topology. This general setting of the network topology unifies most existing models, such as consensus, bipartite consensus, and bipartite containment, which are usually analyzed separately in the same framework using different methods. Relying on a method of constructing cooperative auxiliary digraphs, it is theoretically proved that the agents in closed strong connected components with balanced structure and unbalanced structure gradually reach separately bipartite consensus and consensus, and the agents outside the closed strong connected components gradually enter the convex hull formed by the agents in the closed strong connected components, that is, achieving bipartite containment. Finally, a computer simulation is presented to verify the theoretical discovery.

Aligning naming conventions through dialogue-like interaction in multi-agent systems

Abstract This paper highlights the influence of a dialogue-like interaction pattern on the behaviour of the alignment process in multi-agent Naming Game. In particular, it investigates the difference in performance (iteration- and memory-wise) of the process in baseline one-shot pair-wise interaction pattern against multi-turn pair-wise interaction pattern (similar to dialogue). Presented results indicate that the introduced extension to interaction pattern does affect and shapes the behaviour of the naming-convention alignment process. In particular, iteration-wise multi-turn pattern results in a visible improvement (understood as a significant reduction of needed interactions to reach a coherent state in the population) as compared to baseline one-shot pattern in the original formulation of the Naming Game.

Graphical Nash Equilibria and Replicator Dynamics on Complex Networks.

Pairwise-interaction graphical games have been widely used in the study and design of strategic interaction in multiagent systems. With regard to this issue, one entitative problem is actually to understand how the interaction structure of agents affects the strategy configuration of Nash equilibria. This paper intends to study the effect of interaction networks on Nash equilibria in pairwise-interaction graphical games. We first show that interaction networks may induce new strategy equilibria in pairwise-interaction graphical games and then provide graphical conditions for the existence of these network-induced equilibria. Furthermore, to determine Nash equilibria of pairwise-interaction graphical games, a graphical replicator dynamics model is formulated, and its connection with graphical games is established. In detail, it is shown that every Nash equilibrium of the graphical games corresponds to a fixed point of the graphical replicator dynamics and that every asymptotically stable fixed point of the graphical replicator dynamics corresponds to a strict pure Nash equilibrium of the graphical games. The obtained results are applied in understanding coordination in complex networks and determination of structural conflicts in signed graphs. This work may provide new insights into understanding and designing strategy equilibria and dynamics in games on networks.

The early bird catches the worm: First verify, then monitor!

Trace expressions are a compact and expressive formalism, initially devised for runtime verification of agent interactions in multiagent systems, which has been successfully employed to model real-world protocols, and to generate monitors for mainstream multiagent system platforms, and generalized to support runtime verification of different kinds of properties and systems. In this paper, we propose an algorithm to check Linear Temporal Logic (LTL) properties satisfiability on trace expressions. To do this, we show how to translate a trace expression into a Büchi Automaton in order to realize an Automata-Based Model Checking . We show that this translation generates an over-approximation of our trace expression leading us to obtain a sound procedure to verify LTL properties. Once we have statically checked a set of LTL properties, we can conclude that: (1) our trace expression is formally correct (2) since we use this trace expression to generate monitors checking the runtime behavior of the system, the LTL properties verified by this trace expression are also verified by the monitored system. • Proposal of a static/dynamic hybrid approach to improve runtime monitoring. • Model Checking of a formal specification used to verify object-oriented programs at runtime. • Definition and experimentation of a 3-Steps algorithm to solve the problem.

Specifying autonomy in the Internet of Things: the autonomy model and notation

Driven by digitization in society and industry, automating behavior in an autonomous way substantially alters industrial value chains in the smart service world. As processes are enhanced with sensor and actuator technology, they become digitally interconnected and merge into an Internet of Things (IoT) to form cyber-physical systems (CPS). Using these automated systems, enterprises can improve the performance and quality of their operations. However, currently it is neither feasible nor reasonable to equip any machine with full autonomy when networking with other machines or people. It is necessary to specify rules for machine behavior that also determine an adequate degree of autonomy to realize the potential benefits of the IoT. Yet, there is a lack of methodologies and guidelines to support the design and implementation of machines as explicit autonomous agents such that many designs only consider autonomy implicitly. To address this research gap, we perform a comprehensive literature review to extract 12 requirements for the design of autonomous agents in the IoT. We introduce a set of constitutive characteristics for agents and introduce a classification framework for interactions in multi-agent systems. We integrate our findings by developing a conceptual modeling language consisting of a meta model and a notation that facilitates the specification and design of autonomous agents within the IoT as well as CPS: the Autonomy Model and Notation. We illustrate and discuss the approach and its limitations.

Linear and nonlinear stochastic distribution for consensus problem in multi-agent systems

This paper presents a linear and nonlinear stochastic distribution for the interactions in multi-agent systems (MAS). The interactions are considered for the agents to reach a consensus using hetero-homogeneous transition stochastic matrices. The states of the agents are presented as variables sharing information in the MAS dynamically. The paper studies the interaction among agents for the attainment of consensus by limit behavior from their initial states’ trajectories. The paper provides a linear distribution of DeGroot model compared with a nonlinear distribution of change stochastic quadratic operators (CSQOs), doubly stochastic quadratic operators (DSQOs) and extreme doubly stochastic quadratic operators (EDSQOs) for a consensus problem in MAS. The comparison study is considered for stochastic matrix (SM) and doubly stochastic matrix (DSM) cases of the hetero-homogeneous transition stochastic matrices. In the case of SM, the work’s results show that the DeGroot linear model converges to the same unknown limit while CSQOs, DSQOs and EDSQOs converge to the center. However, the results show that the linear of DeGroot and nonlinear distributions of CSQOs, DSQOs and EDSQOs converge to the center with DSM. Additionally, the case of DSM is observed to converge faster compared to that of SM in the case of nonlinear distribution of CSQOs, DSQOs and EDSQOs. In general, the novelty of this study is in showing that the nonlinear stochastic distribution reaches a consensus faster than all cases. In fact, the EDSQO is a very simple system compared to other nonlinear distributions.

Efficiently detecting switches against non-stationary opponents

Interactions in multiagent systems are generally more complicated than single agent ones. Game theory provides solutions on how to act in multiagent scenarios; however, it assumes that all agents will act rationally. Moreover, some works also assume the opponent will use a stationary strategy. These assumptions usually do not hold in real world scenarios where agents have limited capacities and may deviate from a perfect rational response. Our goal is still to act optimally in these cases by learning the appropriate response and without any prior policies on how to act. Thus, we focus on the problem when another agent in the environment uses different stationary strategies over time. This will turn the problem into learning in a non-stationary environment, posing a problem for most learning algorithms. This paper introduces DriftER, an algorithm that (1) learns a model of the opponent, (2) uses that to obtain an optimal policy and then (3) determines when it must re-learn due to an opponent strategy change. We provide theoretical results showing that DriftER guarantees to detect switches with high probability. Also, we provide empirical results showing that our approach outperforms state of the art algorithms, in normal form games such as prisoner's dilemma and then in a more realistic scenario, the Power TAC simulator.

Integrating social power into the decision-making of cognitive agents

Social power is a pervasive feature with acknowledged impact in a multitude of social processes. However, despite its importance, common approaches to social power interactions in multi-agent systems are rather simplistic and lack a full comprehensive view of the processes involved. In this work, we integrated a comprehensive model of social power dynamics into a cognitive agent architecture based on an operationalization of different bases of social power inspired by theoretical background research in social psychology. The model was implemented in an agent framework that was subsequently used to generate the behavior of virtual characters in an interactive virtual environment. We performed a user study to assess users' perceptions of the agents and found evidence supporting both the social power capabilities provided by the model and their value for the creation of believable and interesting scenarios. We expect that these advances and the collected evidence can be used to support the development of agent systems with an enriched capacity for social agent simulation.

Economic reasoning and artificial intelligence.

The field of artificial intelligence (AI) strives to build rational agents capable of perceiving the world around them and taking actions to advance specified goals. Put another way, AI researchers aim to construct a synthetic homo economicus, the mythical perfectly rational agent of neoclassical economics. We review progress toward creating this new species of machine, machina economicus, and discuss some challenges in designing AIs that can reason effectively in economic contexts. Supposing that AI succeeds in this quest, or at least comes close enough that it is useful to think about AIs in rationalistic terms, we ask how to design the rules of interaction in multi-agent systems that come to represent an economy of AIs. Theories of normative design from economics may prove more relevant for artificial agents than human agents, with AIs that better respect idealized assumptions of rationality than people, interacting through novel rules and incentive systems quite distinct from those tailored for people.

A First-Order Formalization of Commitments and Goals for Planning

Commitments help model interactions in multiagent systems in a computationally realizable yet high-level manner without compromising the autonomy and heterogeneity of the member agents. Recent work shows how to combine commitments with goals and apply planning methods to enable agents to determine their actions. However, previous approaches to modeling commitments are confined to propositional representations, which limits their applicability in practical cases. We propose a first-order representation and reasoning technique that accommodates templatic commitments and goals that may be applied repeatedly with differing bindings for domain objects. Doing so not only leads to a more perspicuous modeling, but also supports many practical patterns.