Downloads & Free Reading Options - Results

In this paper, we study a convergence condition for asynchronous consensus problems in multi-agent systems. The convergence in this context implies the asynchronous consensus value converges to the synchronous one and is unique. Although it is reported in the literature that the consensus value under asynchronous communications may not coincide with the synchronous consensus value, it has not received much attention. In some applications, the discrepancy between them may result in serious consequences. For such applications it is critical to determine under what conditions the asynchronous consensus value is the same as the synchronous consensus value. We illustrate these issues with a few examples and then provide a condition, which guarantees that the asynchronous consensus value converges to the synchronous one. The validity of the proposed result is verified with simulations.

A multi-agent system operates in an uncertain environment about which agents have different and time varying beliefs that, as time progresses, converge to a common belief. A global utility function that depends on the realized state of the environment and actions of all the agents determines the system's optimal behavior. We define the asymptotically optimal action profile as an equilibrium of the potential game defined by considering the expected utility with respect to the asymptotic belief. At finite time, however, agents have not entirely congruous beliefs about the state of the environment and may select conflicting actions. This paper proposes a variation of the fictitious play algorithm which is proven to converge to equilibrium actions if the state beliefs converge to a common distribution at a rate that is at least linear. In conventional fictitious play, agents build beliefs on others' future behavior by computing histograms of past actions and best respond to their expected payoffs integrated with respect to these histograms. In the variations developed here histograms are built using knowledge of actions taken by nearby nodes and best responses are further integrated with respect to the local beliefs on the state of the environment. We exemplify the use of the algorithm in coordination and target covering games.

This paper studies output feedback consensus protocol design problems for linear multi-agent systems with directed graphs. We consider both leaderless and leader-follower consensus with a leader whose control input is nonzero and bounded. We propose a novel sequential observer design approach, which makes it possible to design fully distributed adaptive output feedback protocols that the existing methods fail to accomplish. With the sequential observer architecture, we show that leaderless consensus can be achieved for any strongly connected directed graph in a fully distributed manner, whenever the agents are stabilizable and detectable. For the case with a leader of bounded control input, we further present novel distributed adaptive output feedback protocols, which include nonlinear functions to deal with the effect of the leaders's nonzero control input and are able to achieve leader-follower consensus for any directed graph containing a directed spanning tree with the leader as the root.

Mobile agents are distributed programs which can move autonomously in a network, to perform tasks on behalf of user. They are susceptible to failures due to faults in communication channels, processors or malicious programs. In order to gain solid foundation at the heart of today’s e- society, the mobile agent technology must address the issue of fault tolerance. Checkpointing has been widely used technique for providing fault tolerance in mobile agent systems. But the traditional message passing based checkpointing and rollback algorithms suffer from problems of excess bandwidth consumption and large overheads. This paper proposes use of antecedence graphs and message logs for maintaining fault tolerance information of agents. For checkpointing, dependent agents are marked out using antecedence graphs; and only these agents are involved in process of taking checkpoints. In case of failures, the antecedence graphs and message logs are regenerated for recovery and then normal operation continued. The proposed scheme reports less overheads, speedy execution and reduced recovery times as compared to existing graph based schemes.

For an uncontrollable system, adding leaders and adjusting edge weights are two methods to improve controllability. In this paper, controllability of multi-agent systems under directed topologies is studied, especially on leader selection problem and weight adjustment problem. For a given system, necessary and sufficient algebraic conditions for controllability with fewest leaders are proposed. From another perspective, when leaders are fixed, controllability could be improved by adjusting edge weights, and therefore the system is supposed to be structurally controllable, which holds if and only if the communication topology contains a spanning tree. It is also proved that the number of fewest edges needed to be assigned on new weights equals the rank deficiency of controllability matrix. An algorithm on how to perform weight adjustment is presented. Simulation examples are provided to illustrate the theoretical results.

This paper considers the cooperative output regulation problem for linear multi-agent systems with a directed communication graph, heterogeneous linear subsystems, and an exosystem whose output is available to only a subset of subsystems. Both the cases with nominal and uncertain linear subsystems are studied. For the case with nominal linear subsystems, a distributed adaptive observer-based controller is designed, where the distributed adaptive observer is implemented for the subsystems to estimate the exogenous signal. For the case with uncertain linear subsystems, the proposed distributed observer and the internal model principle are combined to solve the robust cooperative output regulation problem. Compared with the existing works, one main contribution of this paper is that the proposed control schemes can be designed and implemented by each subsystem in a fully distributed fashion for general directed graphs. For the special case with undirected graphs, a distributed output feedback control law is further presented.

In this paper, we first present an adaptive distributed observer for a discrete-time leader system. This adaptive distributed observer will provide, to each follower, not only the estimation of the leader's signal, but also the estimation of the leader's system matrix. Then, based on the estimation of the matrix S, we devise a discrete adaptive algorithm to calculate the solution to the regulator equations associated with each follower, and obtain an estimated feedforward control gain. Finally, we solve the cooperative output regulation problem for discrete-time linear multi-agent systems by both state feedback and output feedback adaptive distributed control laws utilizing the adaptive distributed observer.

In this paper, we formulate and solve a randomized optimal consensus problem for multi-agent systems with stochastically time-varying interconnection topology. The considered multi-agent system with a simple randomized iterating rule achieves an almost sure consensus meanwhile solving the optimization problem $\min_{z\in \mathds{R}^d}\ \sum_{i=1}^n f_i(z),$ in which the optimal solution set of objective function $f_i$ can only be observed by agent $i$ itself. At each time step, simply determined by a Bernoulli trial, each agent independently and randomly chooses either taking an average among its neighbor set, or projecting onto the optimal solution set of its own optimization component. Both directed and bidirectional communication graphs are studied. Connectivity conditions are proposed to guarantee an optimal consensus almost surely with proper convexity and intersection assumptions. The convergence analysis is carried out using convex analysis. We compare the randomized algorithm with the deterministic one via a numerical example. The results illustrate that a group of autonomous agents can reach an optimal opinion by each node simply making a randomized trade-off between following its neighbors or sticking to its own opinion at each time step.

In this paper, we consider the consensus problem of hybrid multi-agent system. First, the hybrid multi-agent system is proposed which is composed of continuous-time and discrete-time dynamic agents. Then, three kinds of consensus protocols are presented for hybrid multi-agent system. The analysis tool developed in this paper is based on the matrix theory and graph theory. With different restrictions of the sampling period, some necessary and sufficient conditions are established for solving the consensus of hybrid multi-agent system. The consensus states are also obtained under different protocols. Finally, simulation examples are provided to demonstrate the effectiveness of our theoretical results.

The consensus problem for multi-agent systems with quantized communication or sensing is considered. Centralized and distributed self-triggered rules are proposed to reduce the overall need of communication and system updates. It is proved that these self-triggered rules realize consensus exponentially if the network topologies have a spanning tree and the quantization function is uniform. Numerical simulations are provided to show the effectiveness of the theoretical results.

Strand spaces are a popular framework for the analysis of security protocols. Strand spaces have some similarities to a formalism used successfully to model protocols for distributed systems, namely multi-agent systems. We explore the exact relationship between these two frameworks here. It turns out that a key difference is the handling of agents, which are unspecified in strand spaces and explicit in multi-agent systems. We provide a family of translations from strand spaces to multi-agent systems parameterized by the choice of agents in the strand space. We also show that not every multi-agent system of interest can be expressed as a strand space. This reveals a lack of expressiveness in the strand-space framework that can be characterized by our translation. To highlight this lack of expressiveness, we show one simple way in which strand spaces can be extended to model more systems.

Strand spaces are a popular framework for the analysis of security protocols. Strand spaces have some similarities to a formalism used successfully to model protocols for distributed systems, namely multi-agent systems. We explore the exact relationship between these two frameworks here. It turns out that a key difference is the handling of agents, which are unspecified in strand spaces and explicit in multi-agent systems. We provide a family of translations from strand spaces to multi-agent systems parameterized by the choice of agents in the strand space. We also show that not every multi-agent system of interest can be expressed as a strand space. This reveals a lack of expressiveness in the strand-space framework that can be characterized by our translation. To highlight this lack of expressiveness, we show one simple way in which strand spaces can be extended to model more systems.

In this paper we present an integrated formal framework for the specification and analysis of Multi-Agent Systems (MAS). Agents are specified in a synchronous programming language called Secure Operations Language (SOL) which supports the modular development of secure agents. Multi-agent systems are constructed from individual agent modules by using the composition operator of SOL, the semantics of which are guaranteed to preserve certain individual agent properties. The formal semantics and the underlying framework of SOL also serve as the basis for analysis and transformation techniques such as abstraction, consistency checking, verification by model checking or theorem proving, and automatic synthesis of agent code. Based on this framework, we are currently developing a suite of analysis and transformation tools for the formal specification, analysis, and synthesis of multi-agent systems.

This document provided a review of current intelligent multi-agent technologies. It presents agent frameworks and agent construction toolkits available in the market today. The Navy Web-based Marketplace is a suitable environment for intelligent multi-agent technologies, where a distributed peer-to-peer process is the essential model with inter-agent communications and coordination. Reasoning with rules of thumb and heuristics are currently common practice as commands define their preferences and detailers make job assignments. This report provides recommendations for agent system development, agent framework, inference engines, and agent construction toolkits. A high-level system architecture is proposed to meet Navy Web-based Marketplace functional requirements. There are three major components of the proposal: (1) a web services-based, front-end at the presentation layer that provides friendly access for Sailors, commands, detailers, and counselors; (2) a workflow layer that provides support for business processes of the various user groups; and (3) and intelligent agent layer that provides the framework through which the agents communicate and negotiate for optimal matches between available Navy jobs and Sailors.

This document provided a review of current intelligent multi-agent technologies. It presents agent frameworks and agent construction toolkits available in the market today. The Navy Web-based Marketplace is a suitable environment for intelligent multi-agent technologies, where a distributed peer-to-peer process is the essential model with inter-agent communications and coordination. Reasoning with rules of thumb and heuristics are currently common practice as commands define their preferences and detailers make job assignments. This report provides recommendations for agent system development, agent framework, inference engines, and agent construction toolkits. A high-level system architecture is proposed to meet Navy Web-based Marketplace functional requirements. There are three major components of the proposal: (1) a web services-based, front-end at the presentation layer that provides friendly access for Sailors, commands, detailers, and counselors; (2) a workflow layer that provides support for business processes of the various user groups; and (3) and intelligent agent layer that provides the framework through which the agents communicate and negotiate for optimal matches between available Navy jobs and Sailors.

Product Distribution (PD) theory is a new framework for controlling Multi-Agent Systems (MAS's). First we review one motivation of PD theory, as the information-theoretic extension of conventional full-rationality game theory to the case of bounded rational agents. In this extension the equilibrium of the game is the optimizer of a Lagrangian of the (probability distribution of) the joint stare of the agents. Accordingly we can consider a team game in which the shared utility is a performance measure of the behavior of the MAS. For such a scenario the game is at equilibrium - the Lagrangian is optimized - when the joint distribution of the agents optimizes the system's expected performance. One common way to find that equilibrium is to have each agent run a reinforcement learning algorithm. Here we investigate the alternative of exploiting PD theory to run gradient descent on the Lagrangian. We present computer experiments validating some of the predictions of PD theory for how best to do that gradient descent. We also demonstrate how PD theory can improve performance even when we are not allowed to rerun the MAS from different initial conditions, a requirement implicit in some previous work.

The design of reactive systems must comply with logical correctness (the system does what it is supposed to do) and timeliness (the system has to satisfy a set of temporal constraints) criteria. In this paper, we propose a global approach for the design of adaptive reactive systems, i.e., systems that dynamically adapt their architecture depending on the context. We use the timed automata formalism for the design of the agents' behavior. This allows evaluating beforehand the properties of the system (regarding logical correctness and timeliness), thanks to model-checking and simulation techniques. This model is enhanced with tools that we developed for the automatic generation of code, allowing to produce very quickly a running multi-agent prototype satisfying the properties of the model.

The design of reactive systems must comply with logical correctness (the system does what it is supposed to do) and timeliness (the system has to satisfy a set of temporal constraints) criteria. In this paper, we propose a global approach for the design of adaptive reactive systems, i.e., systems that dynamically adapt their architecture depending on the context. We use the timed automata formalism for the design of the agents' behavior. This allows evaluating beforehand the properties of the system (regarding logical correctness and timeliness), thanks to model-checking and simulation techniques. This model is enhanced with tools that we developed for the automatic generation of code, allowing to produce very quickly a running multi-agent prototype satisfying the properties of the model.

The design of reactive systems must comply with logical correctness (the system does what it is supposed to do) and timeliness (the system has to satisfy a set of temporal constraints) criteria. In this paper, we propose a global approach for the design of adaptive reactive systems, i.e., systems that dynamically adapt their architecture depending on the context. We use the timed automata formalism for the design of the agents' behavior. This allows evaluating beforehand the properties of the system (regarding logical correctness and timeliness), thanks to model-checking and simulation techniques. This model is enhanced with tools that we developed for the automatic generation of code, allowing to produce very quickly a running multi-agent prototype satisfying the properties of the model.

This viewgraph presentation provides an overview of past and possible future applications for artifical intelligence (AI) in astronaut instruction and training. AI systems have been used in training simulation for the Hubble Space Telescope repair, the International Space Station, and operations simulation for the Mars Exploration Rovers. In the future, robots such as may work as partners with astronauts on missions such as planetary exploration and extravehicular activities.

The cooperative output regulation problem of linear multi-agent systems was formulated and studied by the distributed observer approach in [20, 21]. Since then, several variants and extensions have been proposed, and the technique of the distributed observer has also been applied to such problems as formation, rendezvous, flocking, etc. In this chapter, we will first present a more general formulation of the cooperative output regulation problem for linear multi-agent systems that includes some existing versions of the cooperative output regulation problem as special cases. Then, we will describe a more general distributed observer. Finally, we will simplify the proof of the main results by more explicitly utilizing the separation principle and the certainty equivalence principle.

This paper presents a framework for automatic synthesis of a control sequence for multi-agent systems governed by continuous linear dynamics under timed constraints. First, the motion of the agents in the workspace is abstracted into individual Transition Systems (TS). Second, each agent is assigned with an individual formula given in Metric Interval Temporal Logic (MITL) and in parallel, the team of agents is assigned with a collaborative team formula. The proposed method is based on a correct-by-construction control synthesis method, and hence guarantees that the resulting closed-loop system will satisfy the specifications. The specifications considers boolean-valued properties under real-time. Extended simulations has been performed in order to demonstrate the efficiency of the proposed controllers.

Results obtained with the grant fall into several groupings: (a) Combinatorial conditions on the graphical representation of a two dimensional sensor network that will guarantee localizability of the network in the event of loss of any p sensors and/or q links in the network, for nonnegative integers p and q; (b) analysis of the effects of measurement error on the quality of localization of sensor positions in a sensor network, or more generally a target being localized; (c) the derivation of a measure, including algorithms for computing it, of the quality of connectivity of a network modeled by a graph with nodes and links, and in which the individual links are operative with defined a priori probabilities, and the probability that any one link is operative is independent of the probability that any other link is operative; (d) connectivity and capacity of networks with randomly positioned nodes and probabilistic channel models; (e) Doppler localization problems and miscellaneous multi-agent problems.

We propose two distributed dynamic triggering laws to solve the consensus problem for multi-agent systems with event-triggered control. Compared with existing triggering laws, the proposed triggering laws involve internal dynamic variables which play an essential role to guarantee that the triggering time sequence does not exhibit Zeno behavior. Some existing triggering laws are special cases of our dynamic triggering laws. Under the condition that the underlying graph is undirected and connected, it is proven that the proposed dynamic triggering laws together with the event-triggered control make the state of each agent converges exponentially to the average of the agents' initial states. Numerical simulations illustrate the effectiveness of the theoretical results and show that the dynamic triggering laws lead to reduction of actuation updates and inter-agent communications.

This paper presents a novel method for controlling teams of unmanned aerial vehicles using Stochastic Optimal Control (SOC) theory. The approach consists of a centralized high-level planner that computes optimal state trajectories as velocity sequences, and a platform-specific low-level controller which ensures that these velocity sequences are met. The planning task is expressed as a centralized path-integral control problem, for which optimal control computation corresponds to a probabilistic inference problem that can be solved by efficient sampling methods. Through simulation we show that our SOC approach (a) has significant benefits compared to deterministic control and other SOC methods in multimodal problems with noise-dependent optimal solutions, (b) is capable of controlling a large number of platforms in real-time, and (c) yields collective emergent behaviour in the form of flight formations. Finally, we show that our approach works for real platforms, by controlling a team of three quadrotors in outdoor conditions.

This paper presents a novel method for controlling teams of unmanned aerial vehicles using Stochastic Optimal Control (SOC) theory. The approach consists of a centralized high-level planner that computes optimal state trajectories as velocity sequences, and a platform-specific low-level controller which ensures that these velocity sequences are met. The planning task is expressed as a centralized path-integral control problem, for which optimal control computation corresponds to a probabilistic inference problem that can be solved by efficient sampling methods. Through simulation we show that our SOC approach (a) has significant benefits compared to deterministic control and other SOC methods in multimodal problems with noise-dependent optimal solutions, (b) is capable of controlling a large number of platforms in real-time, and (c) yields collective emergent behaviour in the form of flight formations. Finally, we show that our approach works for real platforms, by controlling a team of three quadrotors in outdoor conditions.

We develop a dynamic multi-agent model of an interbank payment system where banks choose their level of available funds on the basis of private payoff maximisation. The model consists of the repetition of a simultaneous move stage game with incomplete information, incomplete monitoring, and stochastic payoffs. Adaptation takes place with bayesian updating, with banks maximizing immediate payoffs. We carry out numerical simulations to solve the model and investigate two special scenarios: an operational incident and exogenous throughput guidelines for payment submission. We find that the demand for intraday credit is an S-shaped function of the cost ratio between intraday credit costs and the costs associated with delaying payments. We also find that the demand for liquidity is increased both under operational incidents and in the presence of effective throughput guidelines.

Resource allocation is an important issue in cognitive radio systems. It can be done by carrying out negotiation among secondary users. However, significant overhead may be incurred by the negotiation since the negotiation needs to be done frequently due to the rapid change of primary users' activity. In this paper, a channel selection scheme without negotiation is considered for multi-user and multi-channel cognitive radio systems. To avoid collision incurred by non-coordination, each user secondary learns how to select channels according to its experience. Multi-agent reinforcement leaning (MARL) is applied in the framework of Q-learning by considering the opponent secondary users as a part of the environment. The dynamics of the Q-learning are illustrated using Metrick-Polak plot. A rigorous proof of the convergence of Q-learning is provided via the similarity between the Q-learning and Robinson-Monro algorithm, as well as the analysis of convergence of the corresponding ordinary differential equation (via Lyapunov function). Examples are illustrated and the performance of learning is evaluated by numerical simulations.

Resource allocation is an important issue in cognitive radio systems. It can be done by carrying out negotiation among secondary users. However, significant overhead may be incurred by the negotiation since the negotiation needs to be done frequently due to the rapid change of primary users' activity. In this paper, a channel selection scheme without negotiation is considered for multi-user and multi-channel cognitive radio systems. To avoid collision incurred by non-coordination, each user secondary learns how to select channels according to its experience. Multi-agent reinforcement leaning (MARL) is applied in the framework of Q-learning by considering the opponent secondary users as a part of the environment. The dynamics of the Q-learning are illustrated using Metrick-Polak plot. A rigorous proof of the convergence of Q-learning is provided via the similarity between the Q-learning and Robinson-Monro algorithm, as well as the analysis of convergence of the corresponding ordinary differential equation (via Lyapunov function). Examples are illustrated and the performance of learning is evaluated by numerical simulations.

This paper addresses the distributed consensus problem for a linear multi-agent system with switching directed communication topologies. By appropriately introducing a linear transformation, the consensus problem is equivalently converted to a stabilization problem for a class of switched linear systems. Some sufficient consensus conditions are then derived by using tools from the matrix theory and stability analysis of switched systems. It is proved that consensus in such a multi-agent system can be ensured if each agent is stabilizable and each possible directed topology contains a directed spanning tree. Finally, a numerical simulation is given for illustration.

The questions which we will consider here are ‘What is AI?’ and ‘How can we make AI?’. Here we will present the definition of AI in terms of multi-agent systems. This means that here you will not find a new answer to the question ‘What is AI?’, but an old answer in a new form. This new form of the definition of AI is of interest for the theory of multi-agent systems because it gives us better understanding of this theory. More important is that this work will help us answer the second question. We want to make a program which is capable of constructing a model of its environment. Every multi-agent model is equivalent to a single-agent model but multi-agent models are more natural and accordingly more easily discoverable.

The questions which we will consider here are âWhat is AI?â and âHow can we make AI?â. Here we will present the definition of AI in terms of multi-agent systems. This means that here you will not find a new answer to the question âWhat is AI?â, but an old answer in a new form. This new form of the definition of AI is of interest for the theory of multi-agent systems because it gives us better understanding of this theory. More important is that this work will help us answer the second question. We want to make a program which is capable of constructing a model of its environment. Every multi-agent model is equivalent to a single-agent model but multi-agent models are more natural and accordingly more easily discoverable.

The questions which we will consider here are âWhat is AI?â and âHow can we make AI?â. Here we will present the definition of AI in terms of multi-agent systems. This means that here you will not find a new answer to the question âWhat is AI?â, but an old answer in a new form. This new form of the definition of AI is of interest for the theory of multi-agent systems because it gives us better understanding of this theory. More important is that this work will help us answer the second question. We want to make a program which is capable of constructing a model of its environment. Every multi-agent model is equivalent to a single-agent model but multi-agent models are more natural and accordingly more easily discoverable.

We have pioneered the design of optimal event based remote estimation systems for different types of information constraints. In [2] we investigate the case when multiple sensor estimator pairs must share a resource limited network, while in we formulate and solve the case in which the communication from the sensors to the estimators is hindered by collisions. In [6] we propose methods to design optimal remote estimation systems when channels that are subject to energy harvesting mechanisms are used to disseminate in formation from the sensors to the estimators.

We have pioneered the design of optimal event based remote estimation systems for different types of information constraints. In [2] we investigate the case when multiple sensor estimator pairs must share a resource limited network, while in we formulate and solve the case in which the communication from the sensors to the estimators is hindered by collisions. In [6] we propose methods to design optimal remote estimation systems when channels that are subject to energy harvesting mechanisms are used to disseminate in formation from the sensors to the estimators.

Control of multi-agent systems via game theory is investigated. Assume a system level object is given, the utility functions for individual agents are designed to convert a multi-agent system into a potential game. First, for fixed topology, a necessary and sufficient condition is given to assure the existence of local information based utility functions. Then using local information the system can converge to a maximum point of the system object, which is a Nash equilibrium. It is also proved that a networked evolutionary potential game is a special case of this multi-agent system. Second, for time-varying topology, the state based potential game is utilized to design the optimal control. A strategy based Markov state transition process is proposed to assure the existence of state based potential function. As an extension of the fixed topology case, a necessary and sufficient condition for the existence of state depending utility functions using local information is also presented. It is also proved that using better reply with inertia strategy, the system converges to a maximum strategy of the state based system object, which is called the recurrent state equilibrium.

The design of reactive systems must comply with logical correctness (the system does what it is supposed to do) and timeliness (the system has to satisfy a set of temporal constraints) criteria. In this paper, we propose a global approach for the design of adaptive reactive systems, i.e., systems that dynamically adapt their architecture depending on the context. We use the timed automata formalism for the design of the agents' behavior. This allows evaluating beforehand the properties of the system (regarding logical correctness and timeliness), thanks to model-checking and simulation techniques. This model is enhanced with tools that we developed for the automatic generation of code, allowing to produce very quickly a running multi-agent prototype satisfying the properties of the model.

The design of reactive systems must comply with logical correctness (the system does what it is supposed to do) and timeliness (the system has to satisfy a set of temporal constraints) criteria. In this paper, we propose a global approach for the design of adaptive reactive systems, i.e., systems that dynamically adapt their architecture depending on the context. We use the timed automata formalism for the design of the agents' behavior. This allows evaluating beforehand the properties of the system (regarding logical correctness and timeliness), thanks to model-checking and simulation techniques. This model is enhanced with tools that we developed for the automatic generation of code, allowing to produce very quickly a running multi-agent prototype satisfying the properties of the model.

Complex-valued Laplacians have been shown to be powerful tools in the study of distributed coordination of multi-agent systems in the plane including formation shape control problems and set surrounding control problems. In this paper, we first provide some characterizations of complex Laplacians. As an application, we then establish some necessary and sufficient conditions to ensure that the agents interacting on complex-weighted networks converge to consensus in some sense. These general consensus results are used to discuss some multi-agent coordination problems in the plane.

Complex-valued Laplacians have been shown to be powerful tools in the study of distributed coordination of multi-agent systems in the plane including formation shape control problems and set surrounding control problems. In this paper, we first provide some characterizations of complex Laplacians. As an application, we then establish some necessary and sufficient conditions to ensure that the agents interacting on complex-weighted networks converge to consensus in some sense. These general consensus results are used to discuss some multi-agent coordination problems in the plane.

One key issue in multi-agent systems (MAS) is their ability to interact and exchange information autonomously across applications. To secure agent interoperability, designers must rely on a communication protocol that allows software agents to exchange meaningful information. In this paper we propose using ontologies as such communication protocol. Ontologies capture the semantics of the operations and services provided by agents, allowing interoperability and information exchange in a MAS. Ontologies are a formal, machine processable, representation that allows to capture the semantics of a domain and, to derive meaningful information by way of logical inference. In our proposal we use a formal knowledge representation language (OWL) that translates into Description Logics (a subset of first order logic), thus eliminating ambiguities and providing a solid base for machine based inference. The main contribution of this approach is to make the requirements explicit, centralize the specification in a single document (the ontology itself), at the same that it provides a formal, unambiguous representation that can be processed by automated inference machines.

Intelligent agent technology offers good potential to integrate supply chain processes and other business-to-business exchanges more closely than current Web technologies and without the rigid inflexibility associated with EDI applications. Viewing respective procurement and order fulfillment processes of buyer and seller as an integrated whole along the supply chain, we identify opportunities for virtual supply chain re-intermediation through multi-agent systems. Using a proof-concept agent federation, called the Intelligent Mall, we examine the feasibility and performance implications of this approach for a software supply chain. A long-lived, autonomous agent federation is observed performing delegated supply chain tasks as prescribed and in an intelligent, context-sensitive and policy-conforming manner. The kind of agent-integrated supply chain investigated through this work appears to be inherently scalable yet customizable to the level of a specific individual in the organization. This approach to supply chain re-intermediation also appears to generalize well to a wide variety of corporations, universities and other enterprise forms, and the intelligent mall examined in this report appears to offer good potential for improved process performance. This report summarizes exploratory results from experience with the Intelligent Mall and outlines a number of lessons learned and guidelines for implementing multi-agent systems. The conclusions drawn from this investigation are important for executives and managers considering leading adoption of agent technology, and the research agenda is intended to help guide and stimulate continued work along these lines.

The research carried out during 1994 under this grant expands our previous work on two central design issues in multi-agent systems and begins to develop two new tools for multi-agent systems. As part of our effort to provide foundations for more powerful agent-oriented programming environments (AOP), we have been concentrating on basic understanding of mental states and their dynamics. In order to facilitate efficient interactions in shared multi-agent environments, such as the Internet, we have continued our investigation of computational social laws and their implementation in different domains. In order to help model agents within a multi-agent systems we have developed an approach for ascribing mental qualities to agents. Finally, we have suggested an approach for describing the complexity of robotics domains, which has yielded an algorithm for analyzing the possibility of decentralizing tasks.

The research carried out during 1994 under this grant expands our previous work on two central design issues in multi-agent systems and begins to develop two new tools for multi-agent systems. As part of our effort to provide foundations for more powerful agent-oriented programming environments (AOP), we have been concentrating on basic understanding of mental states and their dynamics. In order to facilitate efficient interactions in shared multi-agent environments, such as the Internet, we have continued our investigation of computational social laws and their implementation in different domains. In order to help model agents within a multi-agent systems we have developed an approach for ascribing mental qualities to agents. Finally, we have suggested an approach for describing the complexity of robotics domains, which has yielded an algorithm for analyzing the possibility of decentralizing tasks.

In previous work, a Cooperative Receding Horizon (CRH) controller was developed for solving cooperative multi-agent problems in uncertain environments. In this paper, we overcome several limitations of this controller, including potential instabilities in the agent trajectories and poor performance due to inaccurate estimation of a reward-to-go function. We propose an event-driven CRH controller to solve the maximum reward collection problem (MRCP) where multiple agents cooperate to maximize the total reward collected from a set of stationary targets in a given mission space. Rewards are non-increasing functions of time and the environment is uncertain with new targets detected by agents at random time instants. The controller sequentially solves optimization problems over a planning horizon and executes the control for a shorter action horizon, where both are defined by certain events associated with new information becoming available. In contrast to the earlier CRH controller, we reduce the originally infinite-dimensional feasible control set to a finite set at each time step. We prove some properties of this new controller and include simulation results showing its improved performance compared to the original one.

Inspired by successful biological collective decision mechanisms such as honey bees searching for a new colony or the collective navigation of fish schools, we consider a mean field games (MFG)-like scenario where a large number of agents have to make a choice among a set of different potential target destinations. Each individual both influences and is influenced by the group's decision, as well as the mean trajectory of all the agents. The model can be interpreted as a stylized version of opinion crystallization in an election for example. The agents' biases are dictated first by their initial spatial position and, in a subsequent generalization of the model, by a combination of initial position and a priori individual preference. The agents have linear dynamics and are coupled through a modified form of quadratic cost. Fixed point based finite population equilibrium conditions are identified and associated existence conditions are established. In general multiple equilibria may exist and the agents need to know all initial conditions to compute them precisely. However, as the number of agents increases sufficiently, we show that 1) the computed fixed point equilibria qualify as epsilon Nash equilibria, 2) agents no longer require all initial conditions to compute the equilibria but rather can do so based on a representative probability distribution of these conditions now viewed as random variables. Numerical results are reported.

In this paper, we present a distributed estimation setup where local agents estimate their states from relative measurements received from their neighbours. In the case of heterogeneous multi-agent systems, where only relative measurements are available, this is of high relevance. The objective is to improve the scalability of the existing distributed estimation algorithms by restricting the agents to estimating only their local states and those of immediate neighbours. The presented estimation algorithm also guarantees robust performance against model and measurement disturbances. It is shown that it can be integrated into output synchronization algorithms.

We consider the problem of detecting norm violations in open multi-agent systems (MAS). We show how, using ideas from scrip systems, we can design mechanisms where the agents comprising the MAS are incentivised to monitor the actions of other agents for norm violations. The cost of providing the incentives is not borne by the MAS and does not come from fines charged for norm violations (fines may be impossible to levy in a system where agents are free to leave and rejoin again under a different identity). Instead, monitoring incentives come from (scrip) fees for accessing the services provided by the MAS. In some cases, perfect monitoring (and hence enforcement) can be achieved: no norms will be violated in equilibrium. In other cases, we show that, while it is impossible to achieve perfect enforcement, we can get arbitrarily close; we can make the probability of a norm violation in equilibrium arbitrarily small. We show using simulations that our theoretical results hold for multi-agent systems with as few as 1000 agents---the system rapidly converges to the steady-state distribution of scrip tokens necessary to ensure monitoring and then remains close to the steady state.

In this paper, we study the consensus problem for continuous-time and discrete-time multi-agent systems in state-dependent switching networks. In each case, we first consider the networks with fixed connectivity, in which the communication between adjacent agents always exists but the influence could possibly become negligible if the transmission distance is long enough. It is obtained that consensus can be reached under a restriction of either the decaying rate of the transmission weight or the initial states of the agents. After then we investigate the networks with state-dependent connectivity, in which the information transmission between adjacent agents gradually vanishes if their distance exceeds a fixed range. In such networks, we prove that the realization of consensus requires the validity of some initial conditions. Finally, the conclusions are applied to models with the transmission law of C-S model, opinion dynamics and the rendezvous problem, the corresponding simulations are also presented.