Distributed AI

Smart manufacturing environments (digitalized production systems with integrated sensor networks and data analytics capabilities) require advanced predictive maintenance capabilities, yet implementation faces significant barriers due to data privacy concerns and proprietary knowledge protection requirements. Traditional machine learning approaches necessitate centralized data repositories, creating obstacles for collaborative maintenance optimization across organizational boundaries. This research develops and evaluates a federated learning framework that enables effective predictive maintenance while preserving data privacy in manufacturing networks. The study implemented a horizontal federated learning architecture with secure aggregation protocols and differential privacy techniques across multiple aerospace manufacturing facilities. System performance was evaluated through comparative analysis against centralized and standalone approaches across multiple predictive maintenance use cases. The federated approach achieved 93.7% of centralized model accuracy while eliminating cross-facility data sharing, with failure prediction lead times approaching centralized performance while substantially outperforming standalone models. Computational overhead increased modestly, but network data transfer requirements decreased by 94%. Privacy analysis confirmed that proprietary process parameters could not be reconstructed from shared model updates. This research advances smart manufacturing capabilities by providing a practical implementation framework for privacy-preserving predictive maintenance across organizational boundaries, enabling industry collaboration while maintaining intellectual property protection.

As cross-border financial transactions grow in scale and complexity, so too does the risk of fraud, regulatory noncompliance, and systemic vulnerabilities in global FinTech ecosystems. The heterogeneous regulatory environments, varying KYC/AML standards, and speed of digital finance innovation challenge traditional surveillance and compliance mechanisms. Conventional rule-based fraud detection systems often fall short in adapting to rapidly evolving threat patterns, particularly in high-volume, real-time cross-border contexts. This study introduces a novel framework based on multi-agent artificial intelligence (AI) systems designed to enhance fraud surveillance, increase transparency, and ensure regulatory compliance in global FinTech operations. The proposed architecture comprises autonomous, cooperative AI agents-each specialized in tasks such as behavioral profiling, transaction risk scoring, anomaly detection, and jurisdiction-specific regulation enforcement. These agents operate across decentralized data environments while maintaining privacy and interoperability through secure federated learning protocols. The paper explores how the multi-agent framework dynamically integrates data from diverse sources including digital wallets, blockchain ledgers, and SWIFT/ISO 20022 messaging formats. Agents leverage machine learning models for adaptive fraud pattern recognition and use explainable AI (XAI) to ensure decision traceability. Regulatory compliance agents monitor evolving legal requirements, generating automated audit trails to facilitate international supervisory reporting and minimize compliance latency. Case studies involving real-time remittance flows and digital asset transfers are used to evaluate the system's efficacy in mitigating fraud and false positives. The results demonstrate improved detection accuracy, faster resolution times, and enhanced trust between institutions and regulators. By deploying multi-agent AI, FinTech platforms can achieve secure, transparent, and compliant surveillance in the complex terrain of global financial exchange.

Many reports estimated that in 2024, the number of Internet of Things (IoT) devices exceeded 18 billion worldwide, with predictions suggesting that it could reach nearly 40 billion by 2033. Despite primarily being consumer devices, a growing number of them will find use in industrial and enterprise applications. This shows the significance of IoT and how it shapes the future. However, to realize its full potential, we must address its emerging challenges and highlight recent applications, advances, and trends, which is the focus of this book. The main objective of this book is to capture the state of the art in IoT. This peer-reviewed book serves as a reference for researchers, academics, practitioners, and graduate-level students.

Edge AI represents a transformative shift in artificial intelligence deployment, moving computational intelligence from centralized cloud infrastructure to distributed edge devices and servers. This paradigm evolution addresses critical challenges in latency, bandwidth utilization, and real-time processing requirements that modern applications demand. By leveraging optimized models and specialized hardware

This thesis presents a connectionist architecture, which is able to learn how to play the popular Swedish game ”Brio Labyrinth Game”. The aim of the game is to manoeuvre a steel ball through a complex maze with holes and walls by tipping it in two planar directions using two knobs. The thesis has two main goals. The first goal is to simulate the game using ODE (Open Dynamics Engine) [23] that allows a realistic reproduction of the physical properties of the game. The second is the design and implementation of a biologically inspired agent, which is able to learn how to play the game autonomously. The proposed solution combines the principles of neural networks and reinforcement learning. To learn the proper movements and to deal with the convergence difficulty of reinforcement learning in a continuous state-space representation, a hierarchical learning architecture based on the connectionist Q-learning framework QCON [10] has been developed. As observed in human players and followin...

Most work done in distributed artificial intelligence (DAI) had targeted sensory networks, including air traffic control, urban traffic control, and robotic systems. The main reason is that these applications necessitate distributed interpretation and distributed planning by means of intelligent sensors. Planning includes not only the activities to be undertaken, but also the use of material and cognitive resources to accomplish interpretation tasks and planning tasks. These application areas are also characterized by a natural distribution of sensors and receivers in space. In other words, the sensory data-interpretation tasks and action planning are inter-dependent in time and space. For example, in air traffic control, a plan for guiding an aircraft must be coordinated with the plans of other nearby aircraft to avoid collisions.

Manufacturing enterprises are now moving towards open architectures for integrating their activities with those of their suppliers, customers and partners within wide supply chain networks. Agent-based technology provides a natural way to design and implement such environments. This paper presents an agent-based manufacturing enterprise infrastructure. After a brief review of recent advancements in this domain, we describe the main features of the proposed infrastructure and the functions of its components. A machine-centered dynamic scheduling/rescheduling mechanism is then detailed and a prototype implementation is presented.

In this research-in-progress paper we present a new real world domain for studying the aggregation of different opinions: early stage architectural design of buildings. This is an important real world application, not only because building design and construction is one of the world's largest industries measured by global expenditures, but also because the early stage design decision making has a significant impact on the energy consumption of buildings. We present a mapping between the domain of architecture and engineering research and that of the agent models present in the literature. We study the importance of forming diverse teams when aggregating the opinions of different agents for architectural design, and also the effect of having agents optimizing for different factors of a multi-objective optimization design problem. We find that a diverse team of agents is able to provide a higher number of top ranked solutions for the early stage designer to choose from. Finally, we present the next steps for a deeper exploration of our questions.

This paper presents the results of the R6seau rut6 (smart net) project, the goal of which is to use distributed AI and multi-agent techniques for network management and supervision. More precisely, these techniques have been applied to the partial automation of the dynamic processing (what is known about a network is always incomplete and can change at any time) of alarms and of various event notifications received by network management platforms. The system that we propose is able for example to automatically handle some alarms or to filter events of no-interest for a given operator. To achieve this goal, an assistant, or interface agent according to the model proposed by Patti Maes [MK93], has been realized. The goal of the assistant is first to learn, by observation, the behavior of the network supervision operator and second to reproduce such a behavior when the conditions in which the behavior has been learned are detected again. The learned information are stored using chronicles [Gha94]. A chronicle is a data-structure allowing programmers to represent sequences of events while taking temporal knowledge into account. Our assistant has been implemented and tested within Magenta which is a program, written in Smalltalk, that simulates (in a simplified way) a network management platform. This program respects roughly the GDMO and CMIS standards.

This paper presents the results of the R6seau rut6 (smart net) project, the goal of which is to use distributed AI and multi-agent techniques for network management and supervision. More precisely, these techniques have been applied to the partial automation of the dynamic processing (what is known about a network is always incomplete and can change at any time) of alarms and of various event notifications received by network management platforms. The system that we propose is able for example to automatically handle some alarms or to filter events of no-interest for a given operator. To achieve this goal, an assistant, or interface agent according to the model proposed by Patti Maes [MK93], has been realized. The goal of the assistant is first to learn, by observation, the behavior of the network supervision operator and second to reproduce such a behavior when the conditions in which the behavior has been learned are detected again. The learned information are stored using chronicles . A chronicle is a data-structure allowing programmers to represent sequences of events while taking temporal knowledge into account. Our assistant has been implemented and tested within Magenta which is a program, written in Smalltalk, that simulates (in a simplified way) a network management platform. This program respects roughly the GDMO and CMIS standards.

Dans ce papier, nous proposons une nouvelle approche pour calculer un strong backdoor pour des formules mises sous forme normale conjonctive (CNF). Elle est basée sur une utilisation originale d'une méthode de recherche locale qui fournit un renommage maximisant la sous-formule horn-renommable d'une CNF donnée. Plus précisément,à chaqueétape, on choisit de renommer la variable qui fait le plus diminuer le nombre de clauses non-horn. S'il ne reste plus de clauses strictements positives(ou strictement négatives) ou de clauses non-horn dans la formule, notre méthode répond au problème de satisfaisabilité de la formule originale; sinon, on utilise la plus petite sous-formule qui ne soit pas de horn pour en extraire un ensemble de variables (strong backdoor) tel qu'une fois ces variables instanciées, le reste du problème appartientà une classe polynômiale. Les premiers résultatséxpérimentaux montrent que notre approche est prometteuse sur un grand nombre d'instances SAT.

If agents are able to exploit the resources available in a multi-agent domain they must make use of other agents to help them in their tasks. In order to do this it is important that we first of all have an understanding of agency, and then of how goals are generated and subsequently adopted by other agents. In this paper we build upon a three-tiered hierarchy which has been constructed to define objects, agents and autonomous agents, where all autonomous agents are agents, and all agents are objects. In this hierarchy, ...

The Management Systems of Road Transport (MSRT) must include responsibility for planning the routes and schedules of vehicles fleet involved in the road haulage, distribution and logistics. It must ensure that all operations are carried out in maximum safety, environmental controls and traffic congestion, driver hours, customs requirements, and minimum cost. The complexity of the real-time scheduling of transport orders which comes in an asynchronous and dynamic way makes the MSRT especially suitable for using techniques from Distributed AI. To manage this complex field addressed under a high degree of dynamism and uncertainty which is characterized by an inherent distribution of knowledge and control, we propose in this work, a modeling of an MSRT by a multi-agents system, the modeling of the agents and their interaction by AUML language and we deal with the cooperation during tasks planning in the MSRT.

In this paper, we focus on the Coalition Structure Generation (CSG) problem, which involves finding exhaustive and disjoint partitions of agents such that the efficiency of the entire system is optimized. We propose an efficient hybrid algorithm for optimal coalition structure generation called BOSS. When compared to the state-of-the-art, BOSS is shown to perform better by up to 33.63% on benchmark inputs. The maximum time gain by BOSS is 3392 seconds for 27 agents.

Dynamic Programming (DP) is an effective procedure to solve many combinatorial optimization problems and optimal Coalition Structure generation (CSG) is one of them. Optimal CSG is an important combinatorial optimization problem with significant real-life applications. The current best result in terms of worst case time complexity is O(3 n). So, there is a need to find speedy approaches. This paper proposes a parallel dynamic programming algorithm for optimal CSG. We performed the comparison of our algorithm with the DP for CSG problem which happens to be a sequential procedure. The theoretical, as well as the empirical results show that our proposed method works faster than its sequential counterpart. We obtain a speed-up that is almost 14 times in case of 17 agents using a 16-core machine.

Coalition Structure Generation (CSG) is an NP-complete problem that remains difficult to solve on account of its complexity. In this paper, we propose an efficient hybrid algorithm for optimal coalition structure generation called ODSS. ODSS is a hybrid version of two previously established algorithms IDP (Rahwan and Jennings 2008) and IP (Rahwan et al. 2009). ODSS minimizes the overlapping between IDP and IP by dividing the whole search space of CSG into two disjoint sets of subspaces and proposes a novel subspace shrinking technique to reduce the size of the subspace searched by IP with the help of IDP. When compared to the state-of-the-art against a wide variety of value distributions, ODSS is shown to perform better by up to 54.15% on benchmark inputs.

The Coalition Structure Generation (CSG) problem is a partitioning of a set of agents into exhaustive and disjoint coalitions to maximize social welfare. This NP-complete problem arises in many practical scenarios. Prominent examples are included in the field of transportation, e-Commerce, distributed sensor networks, and others. The fastest exact algorithm to solve the CSG problem is ODP-IP, which is a hybrid version of two previously established algorithms, namely Improved Dynamic Programming (IDP) and IP. In this paper, we show that the ODP-IP algorithm performs many redundant operations. To improve ODP-IP, we propose a faster abortion mechanism to speed up IP’s search. Our abortion mechanism decides at runtime which of the IP's operations are redundant to skip them. Then, we propose a modified version of IDP (named MIDP) and an improved version of IP (named IIP). Based on these two improved algorithms, we develop a hybrid version (MIDP-IIP) to solve the CSG problem. After a ...

In this paper, we develop two methods for improving the performance of the standard Distributed Breakout Algorithm (Yokoo et al. 1996) using the notion of interchangeability. We study the performance of this algorithm on the problem of distributed sensor networks. In particular, we consider how neighborhood interchangeability and neighborhood partial interchangeability (Freuder 1991) can be used to keep conflicts localized and avoid "chain reactions" where a conflict originating in one part of the problem spreads to neighboring areas. We see from the experimental results that such techniques can bring about significant improvements in terms of the number of cycles required to solve the problem (and therefore improvements in terms of communication and time requirements), especially for difficult problems. Moreover, the improved algorithms are able to solve a higher proportion of the test problems.

This paper presents an architecture for the control of autonomus agents that allows explicit cooperation among them. The structure of the software agents controlling the robots is based on a general purpose multi-agent architecture based on a two level approach. One level is composed of reactive skills capable of achieving simple actions by their own. The other is based on an agenda used as an opportunistic planning mechanism to compound, activate and coordinate the basic skills. This agenda handles actions both from the internal goals of the robot or from other robots. This paper describes the work already accomplished, as well as the issues arising from the implementation of the architecture and its use in the RoboCup domain.

Nowadays, we can see an increasing amount of robotic systems for pur- poses of continuously growing complexity. Some of these applications require, not just a single robot, but a group or team of robots that must work together to accom- plish the goal. This raises some problems related to the coordination of several robots that must be handled accordingly. Robotics has improved in many ways. Mechanics are becoming faster, more pre- cise, reliable and durable. Electronics provides greater autonomy, reliable sensors, and faster processing. Artificial vision endows robots with the ability to detect objects in the world by just looking at them with video cameras, with no special extra sensors. But, even with all this great new technology, a big problem remains: how to decide what the robot should do? A robotic arm can pick up a glass of water without spilling it, but when should it do that? Why should it do that? Will picking up the glass help to accomplish the goal? This is why the lin...

In this paper we discuss the use of the social concept 'conviviality' for computer science in general, and for the development of ambient technologies in particular. First, we give a survey of the use of the concept 'conviviality' in the social sciences. Conviviality is usually considered a positive concept related to sociability. However, further analysis reveals a negative side related to lack of diversity, privacy and ethical issues. Second, we argue that conviviality requirements for ambient intelligence are challenging, because ambient technologies give rise to a new virtual and social reality, and conviviality issues play a central role in applications that are concerned with the interaction of material, virtual and social realities. Conviviality highlights an important challenge that we illustrate with examples that emphasize ethical issues, such as privacy threats, surveillance of users and identity theft. Intelligent interfaces, for example, allow instant interactions and thereby create strong needs for coordination and regulation mechanisms that have to be addressed to ensure the safeguard of individuals against abuses, such as privacy intrusions and identity manipulations. Third, we propose a conviviality ontology by operationalizing the fuzzy concept of 'conviviality,' such that it can be used in computer science in the same way as other social concepts such as 'service,' 'contract' or 'trust' are used in this area. Conviviality is defined using dependence networks, and tools for conviviality are based on, what we call, conviviality masks. Fourth, we illustrate how convivial ambient intelligence applications can be designed using our operationalized concept of conviviality. We illustrate our arguments and contributions with a running example on the use of ambient technologies in digital cities, as a prototypical example where material reality such as ambient technologies interacts with virtual and social realities.

Immune systems of live forms have been an abundant source of inspiration to contemporary computer scientists. Problem solving strategies, stemming from known immune system phenomena, have been successfully applied to challenging problems of modern computing. However, research in artificial immune systems has overlooked establishing a coherent model of known immune system behaviour. This paper aims reports on an preliminary computer model of an immune system, where each immune system component is specified in terms of its observable behaviour using a suitable process algebra. Our model is not only suitable to simulation but also and more importantly to formal analyzes of immune system behaviour.

Article received on March 27, 2003, accepted on June 08, 2004 Immune systems of live forms have been an abundant source of inspiration to contemporary computer scientists. Problem solving strategies, stemming from known immune system phenomena, have been successfully applied to challenging problems of modern computing. However, research in artificial immune systems has overlooked establishing a coherent model of known immune system behaviour. This paper aims reports on an preliminary computer model of an immune system, where each immune system component is specified in terms of its observable behaviour using a suitable process algebra. Our model is not only suitable to simulation but also and more importantly to formal analyzes of immune system behaviour.

In this paper, we propose a new algorithm for the Coalition Structure Generation (CSG) problem that can be run with more than 28 agents while using a complete set of coalitions as input. The current state-of-the-art limit for exact algorithms to solve the CSG problem within a reasonable time is 27 agents. Our algorithm uses a novel representation of the search space and a new code-based search technique. We propose an effective heuristic search method to efficiently explore the space of coalition structures using our code-based technique and show that our method outperforms existing state-of-the-art algorithms by multiple orders of magnitude.

In this paper, we propose a novel algorithm to address the Coalition Structure Generation (CSG) problem. Specifically, we use a novel representation of the search space that enables it to be explored in a new way. We introduce an index-based exact algorithm. Our algorithm is anytime, produces optimal solutions, and can be run on large-scale problems with hundreds of agents. Our experimental evaluation on a benchmark with several value distributions shows that our representation of the search space that we combined with the proposed algorithm provides high-quality results for the CSG problem and outperforms existing state-of-the-art algorithms.

Design imposes a novel social choice problem: using a team of voting agents, maximize the number of optimal solutions; allowing a user to then take an aesthetical choice. In an open system of design agents, team formation is fundamental. We present the first model of agent teams for design. For maximum applicability, we envision agents that are queried for a single opinion, and multiple solutions are obtained by multiple iterations. We show that diverse teams composed of agents with different preferences maximize the number of optimal solutions, while uniform teams composed of multiple copies of the best agent are in general suboptimal. Our experiments study the model in bounded time; and we also study a real system, where agents vote to design buildings.

Immune systems of live forms have been an abundant source of inspiration to contemporary computer scientists. Problem solving strategies, stemming from known immune system phenomena, have been successfully applied to challenging problems of modern computing. However, research in artificial immune systems has overlooked establishing a coherent model of known immune system behaviour. This paper aims reports on an preliminary computer model of an immune system, where each immune system component is specified in terms of its observable behaviour using a suitable process algebra. Our model is not only suitable to simulation but also and more importantly to formal analyzes of immune system behaviour.

In this paper, we propose a new algorithm for the Coalition Structure Generation (CSG) problem that can be run with more than 28 agents while using a complete set of coalitions as input. The current state-of-the-art limit for exact algorithms to solve the CSG problem within a reasonable time is 27 agents. Our algorithm uses a novel representation of the search space and a new code-based search technique. We propose an effective heuristic search method to efficiently explore the space of coalition structures using our code based technique and show that our method outperforms existing state-of-the-art algorithms by multiple orders of magnitude.

This paper describes and discusses the work carried on in the context of the CORTEX project, for the development of adaptive real-time applications in wormhole based systems. The architecture of CORTEX relies on the existence of a timeliness wormhole, called Timely Computing Base (TCB), which we have described in previous papers. Here we focus on the practical demonstration of the wormhole concept, through a demo with two complementary facets. The objective is to illustrate the effectiveness of the concept from a practical, yet rigorous, perspective, which is done with the help of an emulation framework that we present in the paper. Furthermore, the paper also describes two different ways of implementing timeliness wormholes on top of both wired and wireless infrastructures.

We show that agent programming patterns are well expressed in terms of an object oriented layer extended with a generalized inheritance mechanism and independent logic programming based inference engines. Instead of proposing yet another agent programming "model" we simply derive key agent programming patterns as the natural result of a set of programming language constructs. The suggested equation: Agents = Objects + Logic + Inference Engines + Coordination + Remote Action provides orthogonal agent composition mechanisms which are expressive and highly reusable. The approach described in this paper has emerged in the process of building Agent classes as extensions to our industrial strength Object Oriented Prolog system, Jinni 2002,

The objective of D6 is to design a programming model suitable for the development of proactive applications constructed from mobile sentient objects. D6 embodies the final deliverable and follows its predecessor deliverable D2, the preliminary definition of the CORTEX programming model. It outlines our definition of sentient objects and describes how an application developer may program them in a context aware manner using techniques such as Bayesian networks and belief theory. Furthermore, D6 describes the ...

In this paper, we develop a formalism called a distributed constraint satisfaction prob/em (distributed CSP) and algorithms for solving distributed CSPs. A distributed CSP is a constraint satisfaction problem in which variables and constraints are distributed among multiple agents. Various application problems in Distributed Artificial Intelligence can be formalized as distributed CSPs. We present our newly developed technique called asynchronous backtracking that allows agents to act asynchronously and concurrently without any global control, while guaranteeing the completeness of the algorithm. Furthermore, we describe how the asynchronous backtracking algorithm can be modified into a more efficient algorithm called an asynchronous weak-commitment search, which can revise a bad decision without exhaustive search by changing the priority order of agents dynamically. The experimental results on various example problems show that the asynchronous weak-commitment search algorithm is, ...

Many problems in multi-agent systems can be described as distributed CSPs. However, some real-life problem can be over-constrained and without a set of consistent variable values when described as a distributed CSP. We have presented the distributed partial CSP for handling such an over-constrained situation and the distributed maximal CSP as a subclass of distributed partial CSP. In this paper, we first show another subclass of distributed partial CSP, the distributed hierarchical CSP. Next, we present a series of new algorithms for solving a distributed hierarchical CSP, each of which is designed based on our previous distributed constraint satisfaction algorithms. Finally, we evaluate the performance of our new algorithms on distributed 3coloring problems in terms of optimality and anytime characteristics. The results show that our new algorithms perform much better than the previous algorithm for finding an optimal solution and produce good results for anytime characteristics.

When multiple agents are in a shared environment, there usually exist constraints among the possible actions of these agents. A distributed constraint satisfaction problem (distributed CSP) is a problem to find a consistent combination of actions that satisfies these inter-agent constraints. Various application problems in multi-agent systems can be formalized as distributed CSPs. This paper gives an overview of the existing research on distributed CSPs. First, we briefly describe the problem formalization and algorithms of normal, centralized CSPs. Then, we show the problem formalization and several MAS application problems of distributed CSPs. Furthermore, we describe a series of algorithms for solving distributed CSPs, i.e., the asynchronous backtracking, the asynchronous weak-commitment search, the distributed breakout, and distributed consistency algorithms. Finally, we show two extensions of the basic problem formalization of distributed CSPs, i.e., handling multiple local variables, and dealing with over-constrained problems.

We first present an algorithm called multi-ABT as a baseline algorithm for solving distributed constraint satisfaction problems where each agent has multiple local variables. Then, we show a cost profile of multi-ABT for various numbers of intra-agent constraints (constraints defined over variables of one agent) and inter-agent constraints (constraints defined over variables of multiple agents) in a distributed graph-coloring problem. This cost profile enabled us to make the following observations: (1) the satisfiability thresholds are identified in the narrow region on the x-y plane (where the x-axis is the number of intra-agent constraints and the y-axis is the number of inter-agent constraints) in which the sum of intra-and interagent constraints is constant, and problem instances in the region (called the crossover belt) are likely to be expensive in terms of the median cost; (2) among problem instances on the crossover belt, those with a smaller number of intra-agent constraints and a larger number of inter-agent constraints may be more expensive; and (3) for a fixed total number of variables, problem instances on the crossover belt may be more expensive as the number of agents increases or the number of variables per agent decreases. Our further experiments suggest that these observations can be generalized to cases where different algorithms are applied or different sets of parameters of the problem are used.

In this paper, we propose a new algorithm for the Coalition Structure Generation (CSG) problem that can be run with more than 28 agents while using a complete set of coalitions as input. The current state-of-the-art limit for exact algorithms to solve the CSG problem within a reasonable time is 27 agents. Our algorithm uses a novel representation of the search space and a new code-based search technique. We propose an effective heuristic search method to efficiently explore the space of coalition structures using our code based technique and show that our method outperforms existing state-of-the-art algorithms by multiple orders of magnitude.

Artificial intelligence, on the one hand, and behavioral and social science on the other, are inextricably related because the ability to coordinate in multiagent environments is fundamental to intelligence. Distributed artificial intelligence (DAI) has used several different social metaphors as the basis for building computational theories of intelligent coordination, with significant success. However, the real challenge in developing general theories of coordination is in bridging between the different perspectives implicit in the various social metaphors. In this paper, I describe a line of research that has led to a view of coordination as a distributed search problem, where the search space provides a common representation for organizations, plans, and schedules. The evolving approach is thus a candidate foundation for a truly interdisciplinary study of coordination from both computational and social perspectives.

Determining the optimal coalition structure is an interesting problem in multi-agent system and remains difficult to solve. It is proved that this problem is NPcomplete and even finding a suboptimal solution needs exponential time. The main problem is, after a group of agents comes together, how the agents can be partitioned such that social payoff is maximized? This work presents the application of Algorithm X in coalition structure generation problem. Before the algorithm starts, it uses some kernelization techniques to make the original input (2 − 1) size into some smaller input instances.

Coalition Structure Generation (CSG) is an NP-complete problem that remains difficult to solve on account of its complexity. In this paper, we propose an efficient hybrid algorithm for optimal coalition structure generation called ODSS. ODSS is a hybrid version of two previously established algorithms IDP (Rahwan and Jennings 2008) and IP (Rahwan et al. 2009). ODSS minimizes the overlapping between IDP and IP by dividing the whole search space of CSG into two disjoint sets of subspaces and proposes a novel subspace shrinking technique to reduce the size of the subspace searched by IP with the help of IDP. When compared to the state-of-the-art against a wide variety of value distributions, ODSS is shown to perform better by up to 54.15% on benchmark inputs.

Immune systems of live forms have been an abundant source of inspiration to contemporary computer scientists. Problem solving strategies, stemming from known immune system phenomena, have been successfully applied to challenging problems of modern computing. However, research in artificial immune systems has overlooked establishing a coherent model of known immune system behaviour. This paper aims reports on an preliminary computer

Current models of the immune system have proven capable of reproducing the dynamics of the immune system response. However, they lack of formalisms (including semantics) to understand general laws of immune system behaviour. This paper introduces a process algebra with which we can model the immune system and make this model subjected to formal verification. We shall discuss how to use the model in order to formalise important properties of immune systems and then outline the sorts of mechanisms that are required to formally verify them.

SUMMARYThe use of Multi-Agent Systems as a Distributed AI paradigm for Robotics is the principal aim of our present work. In this paper we consider the needed concepts and a suitable architecture for a set of Agents in order to make it possible for them to cooperate in solving non-trivial tasks.Agents are sets of different software modules, each one implementing a function required for cooperation. A Monitor, an Acquaintance and Self-knowledge Modules, an Agenda and an Input queue, on the top of each Intelligent System, are fundamental modules that guarantee the process of cooperation, while the overall aim is devoted to the community of cooperative Agents. These Agents, which our testbed concerns, include Vision, Planner, World Model and the Robot itself.

Coalition formation in characteristic function games entails agents partitioning themselves into a coalition structure and assigning the numeric rewards of each coalition via a payoff vector. Various coalition structure generation algorithms have been proposed that guarantee that an optimal coalition structure is found. We present the Synergy Coalition Group-based Dynamic Programming (SCG-DP) algorithm that guarantees that an optimal coalition structure and a least core stable payoff vector is found. This is completed by extending the existing results for the Synergy Coalition Group (SCG) representation to show that only coalitions in the SCG are needed to find a weak-least core stable payoff vector. The SCG-DP algorithm builds on this result by performing only the search operations necessary to guarantee that coalitions in the SCG of the given characteristic function game are found. The number of operations required is significantly less for many coalition-value distributions compared to the original Dynamic Programming (DP) algorithm [34] that finds an optimal coalition structure (e.g. only ∼60% of DP's coalition lookup operations are performed in SCG-DP for 18 agents using a normal coalition-value distribution). Our experimental results show that a lower bound for these operations in SCG-DP converges onto 50%. This is an increase on the ∼33% bound of the optimal dynamic programming (ODP) algorithm [14], but ODP does not search for a stable solution.

Complex, real-world domains require a rethinking of traditional approaches to AI planning. Planning and executing the resulting plans in a dynamic environment requires a continual approachinwhich planning and execution are interleaved, there may be uncertaintyin the current and projected world state, and replanning may be required when the situation changes or planned actions fail. Furthermore, complex planning and execution problems may require multiple computational agents and human planners to collaborate on a solution. In this article, we describe a new paradigm for planning in complex, dynamic environments, whichweterm distributed,continual planning (DCP). We argue that developing DCP systems will be necessary in order for planning applications to be successful in these environments. We give a historical overview of research leading up to the current state of the art in DCP, and describe research in distributed and continual planning. The increasing emphasis on r...

Abstract: Recent advances in the area of mobile ad hoc computing and pervasive computing have driven the emergence of new challenges. For example, the “Intelligent Environment” or “Smart Environment” has become one of the key research areas in the pervasive computing arena. Mobile ad hoc scenarios also include time-critical applications such as an autonomous vehicle system in which vehicles are able to operate independently and cooperate with each other to avoid collisions. In this paper, we present a context-aware ...

The activated sludge process-the main biological technology usually applied to wastewater treatment plants (WWTP)-directly depends on live beings (microorganisms), and therefore on unforeseen changes produced by them. It could be possible to get a good plant operation if the supervisory control system is able to react to the changes and deviations in the system and can take the necessary actions to restore the system's performance. These decisions are often based both on physical, chemical, microbiological principles (suitable to be modelled by conventional control algorithms) and on some knowledge (suitable to be modelled by knowledge-based systems). But one of the key problems in knowledge-based control systems design is the development of an architecture able to manage efficiently the different elements of the process (integrated architecture), to learn from previous cases (spec@c experimental knowledge) and to acquire the domain knowledge (general expert knowledge). These problems increase when the process belongs to an ill-structured domain and is composed of several complex operational units. Therefore, an integrated and distributed AI architecture seems to be a good choice. This paper proposes an integrated and distributed supervisory multi-level architecture for the supervision of WWTP, that overcomes some of the main troubles of classical control techniques and those of knowledge-based systems applied to real world systems.

In this paper we propose a distributed object oriented logic programming language, called DK_Parlog++, that we have developed at Imperial College, as a powerful tool for enterprise modelling and for prototyping an enterprise integration framework. We have used the language to develop a model of our own department and to prototype a generic enterprise integration framework based on role theory. We adopt an agent based approach, the agents being essentially distributed active objects containing both procedural methods and public knowledge, the knowledge being expressed as Prolog facts and rules. Each agent is an instance of some agent class. The agents interact with each other by sending messages to request services, and by remote queries to access each others public knowledge. The agents can either be used as a coordination layer for an enterprise integration prototype or to provide an active model of the activities of the enterprise. Each agent inherits all the methods and knowledge of its class and super classes. It has a 'state' partly encoded in its state variables and partly in a dynamic component of its public knowledge. The methods of the agents will be the activity rules, the procedures of the enterprise. The public knowledge of each agent can be used to express the constraints and preconditions of these procedures, or to describe the effects of the procedures. The knowledge, distributed over the the agents and their classes, comprise a distributed 00 knowledge base for the activities of the organisation. This paper assumes acquaintance with the concepts of concurrent object oriented programming and logic programming, particularly Prolog.