Multiagent System Technologies

Agent platforms have been conceived traditionally as middleware, helping to deal with various application challenges like agent programming models, remote messaging, and coordination protocols. A middleware is typically a bundle of functionalities necessary to execute multi-agent applications. In contrast to this traditional view, nowadays different use cases also for selected agent concepts have emerged requiring also different kinds of functionalities. Examples include a platform for conducting multi-agent simulations, intelligent agent behavior models for controlling non-player characters (NPCs) in games and a lightweight version suited for mobile devices. A one-size-fits-all software bundle often does not sufficiently match these requirements, because customers and developers want solutions specifically tailored to their needs, i.e. a small but focused solution is frequently preferred over bloated software with extraneous functionality. Software product lines are an approach suitable for creating a series of similar products from a common code base. In this paper we will show how software product line modeling and technology can help creating tailor-made products from multi-agent platforms. Concretely, the Jadex platform will be analyzed and a feature model as well as an implementation path will be presented.

Autonomous systems, often realized as multi-agent systems, are envisioned to deal with uncertain and dynamic environments. They are applied in dangerous situations, e.g. as rescue robots or to relieve humans from complex and tedious tasks like driving a car or infrastructure maintenance. But in order to further improve the technology a generic measurement and benchmarking of autonomy is required. Within this paper we present an improved understanding of autonomous systems. Based on this foundation we introduce our concept of a multi-dimensional autonomy metric framework that especially takes into account multi-system environments. Finally, our approach is illustrated by means of an example.

While computer simulation gained importance as a technique for generating knowledge in various research disciplines, the size of simulation models representing real world scenarios is growing, too. In Social Simulation, e.g., there is a need to simulate a large number of humans using individual software agents for generating and analyzing human-like behavior in artificial societies. Nowadays, a variety of toolkits and frameworks exists providing functionalities for supporting implementation and execution of simulation experiments. Yet, the choice of a suitable framework is difficult as unforeseen scalability issues may arise when extending agent models. Therefore, this paper aims at providing a method for analyzing and comparing agent-based simulation frameworks regarding their ability to scale simulation models and experiments. Based on performance metrics, standardized experiments are conducted while altering internal and external scaling parameters. As part of the study, four Java-based agent frameworks are analyzed and compared: Aimpulse Spectrum, JADE, MASON, and Repast.

This paper presents the Agent Activity, a Petri net construct combining workflow tasks with the abilities of software agents. Agent Activities are at the core of our new approach, in which they are used by novel modelling entities. These entities provide integrated workflow and agent functionality. This supports a modelling perspective incorporating the strengths of both multi-agent systems and workflow management systems to allow for e.g. better support of inter-organisational modelling. The paper focuses on the role, function and construction of the Agent Activity in the approach.

In this paper a spatio-temporal simulation framework for multiagent systems is introduced. Its fundamental idea consists in the possibility to develop agents that can be easily deployed in different kinds of scenarios without adapting the agents’ percepts, actions or communication model to a specific scenario. This can be useful to observe and evaluate agents in the context of various scenarios, e. g. to measure their generality and adaptivity against different kinds of problems. To demonstrate the framework, two different example scenarios are considered that are both simulated with the same simple agent implementation.

Industry 4.0 scenarios involve Cyber-Physical-Systems to achieve a higher degree of individualization. Multiagent systems show the main characteristics to reach the goal of increased individualization possibilities by flexible interactions of agents. However, the organizational complexity of individualized manufacturing processes and thus the complexity of current supply networks require the extension of current multiagent system models. Enabling interaction between various multiagent systems representing autonomous actors of a supply network is necessary to cope with the increased complexity. This paper presents ongoing research and adds to the literature by modelling multiagent systems as fractals of a supply network using logistics modelling approaches. We present three examples for applying the multiagent perspective to such Industry 4.0 supply networks.

A growing network of technical systems, embedded and autonomous, influence our daily work. Among them, cyber-physical systems establish a close connection between the virtual and the real world. In this paper we show how an existing multiagent system that controls the physical production of goods on a monorail is virtualized by extracting the agents as black boxes and by integrating them into a multiagent simulation system. As a result, the exact same agents run in physical and cyber world. Towards this end, the physical environment has been mapped and visualized. Experiments show that the modeling and simulation error is small, such that scenarios can be varied, tested, debugged, and scaled, saving huge amounts of labor.

In the Web of Things (WoT), special communication networks composed of sensor nodes, actuator nodes and service nodes form the basis for new types of web application systems, which are directly connected to the real world via sensors and actuators. We propose a conceptual framework for simulating WoT systems as multi-agent systems where both sensor nodes, actuator nodes and service nodes, as well as other systems in their environment interacting with them (such as other web applications, web services and human users), are modeled and simulated as agents. Our conceptual framework includes an ontology of WoT systems as sensor/actuator systems, and a meta-model for defining an agent-based WoT system simulation language.

Mobile agents’ security is a major challenge for the expansion of their use. Actually, most research works have been interested in integrating mobile agents’ security requirements in the implementation stage. However, the integration of security requirements in the design stage could help towards the development of more secure systems based on mobile agents. In this paper, we are interested to model the security requirements of mobile agents in the design stage in order to protect places from malicious visitor agents. An example from a teleexpertise system is used to illustrate the proposed approach.

We use agent-based simulation to compare the performance of four scheduling policies in youth health care. The policies deploy push/pull and centralized/decentralized concepts. The simulation model represents an authentic business case and is parameterized with actual market data. The model incorporates, among other things, non-stationary Poisson arrival processes, reneging and return mechanisms, and care provider’s client preferences. We have identified that performance measurement in youth health care should not be focused on queue lengths alone, which is presently the case, but should include a case difficulty parameter as well. The simulation results, together with contextual data obtained from stakeholder interviews, indicate that a push strategy with a centralized queue suits the sector best, which is different from the current real-world situation. This policy ensures a higher level of fairness in treatment provision because the care providers are compelled to take their share in treating the difficult and economically less attractive cases. The complexity of the case cannot be captured by current queuing theory methods. Our simulation approach incorporates these complexities, which turn out to be relevant for the scheduling policy decision. We validate the model and strategies using real market data and field expert discussions.

We study voting rules as a promising option for collective decision making in traffic applications. The aim of our work is to compare the suitability of several voting rules for different traffic applications and to tackle problems which arise when applying voting rules in traffic management. Here, we propose a multi-agent based voting architecture for evaluation of the suitability of voting rules. The design of the voting architecture is informed by the requirements from two applications we intend to study. The J-MADeM architecture is adapted for the development of our architecture. We describe the voting theory model we intend to incorporate in the architecture, the initial applications we plan to investigate and the features of the voting architecture. Furthermore, we outline the first simulation we intend to conduct using the voting architecture, focusing on the aspect of iterative winner determination for the committee voting rules Minisum and Minimax Approval.

Local producers and storage units will play a major role in the future electricity grid along with the challenge of sustainability. For this reason, smart grid simulations are needed to forecast the challenges of two-way information and energy flow. The Power Trading Agent Competition (Power TAC) provides an open source simulation platform to enable and verify various smart grid studies from the perspective of sustainability. Besides, an annual competition is hosted in which autonomous agents trade in energy markets and make profits. AgentUDE won the Power TAC 2014 Final as a broker utilizing an adaptive agent. This paper details the trading strategies of AgentUDE and analyzes the tournament.

Due to the increasing number of Dementia patients, the overall costs for caregiving has grown by 32 % between 2002 and 2008. The efficient use of smart decision support systems for managing ambulant care and mobile nursing services that provide professional care for Dementia patients is an important challenge to reduce cost and increase service quality. The optimal allocation of caregiving resources from different mobile nursing service firms to a growing number of Dementia patients, however, is a difficult problem in the healthcare domain. We approach this problem from a multiagent systems perspective by designing and implementing a distributed decision support system that utilizes an auction-based protocol for allocating caregiving resources subject to Dementia-specific service attributes. We demonstrate the usefulness of the proposed protocol by an early stage prototype implementation presenting the system’s proof-of-concept.

Automotive companies tend to apply modular approaches in their product development processes in order to save costs and meet increasingly diversified customer demands. In largely decentralized environments with cross-branded development projects over multiple departments in different sites this modular approach leads to very complex and large data structures. Maintaining consistency and transparency, as well as coordinating information flows in such an environment is a major task which is often accomplished manually. Based on a real world case study, this paper analyzes a key development process: the connection of geometric (geometries) and logistical data (parts). During this time consuming process information carriers (geometries and parts) with independent lifecycles that are maintained by different stakeholders (designer and purchaser) of different departments (and in this scenario even within multiple brands) are linked as these carriers themselves are mutually dependent. This paper then proceeds to model five agent-based architecture variants to support this process. In addition, an algorithm to map geometric and logistical data which aims to relieve the actors involved (regarding the organizational overhead) is outlined.The paper concludes with a comparison of the different agent architecture variants and emphasizes the most promising variants to partly automate the connection of geometric and logistical data.

Renewable energy sources such as wind and solar are difficult to balance for the grid because they are weather dependent. We study how the storage of electric vehicles can balance a grid with an increasing intermittent renewable energy content in the short term and contribute to a more efficient and sustainable smart grid. In the Power Trading Agent Competition, a mulitagent platform, we represent fleets of electric vehicles to make a tradeoff between the conflicting interests of storing intermittent energy in the electric vehicles and driving them. The richness of this platform allows us to draw conclusions for the future of a sustainable grid with electric vehicles.

Opportunism is a social behavior that achieves own gains at the expense of others. In this study, we propose a formal model of opportunism, which consists of the properties knowledge asymmetry, value opposition and intention, based on situation calculus.

The era of mass customization of goods forces manufacturing systems to promote agility, flexibility and responsiveness, leading to complex and unpredictable systems. Such challenges have an impact in terms of the system responsiveness and adaptation, production costs, product quality, etc. In order to improve those aspects, some flexible control manufacturing paradigms were proposed offering elasticity to change available skills and provide new services. However, the understanding of when and how to (self-) reconfigure the system aiming to perform a fast changeover, is a crucial issue. This work proposes a self-organizing multi-agent system approach for an efficient and on the fly reconfiguration of services in the manufacturing domain. Besides self-organizing techniques, other dimensions, e.g., “social-based” trust and QoS metrics, are used to ensure a constant QoS in an agile production system. The insertion of intelligent agents facilitates the improvement of strategies that perform the service reconfiguration, and in addition, permits to understand when and how self-reconfiguration takes place in order to allow a continuous improvement of the system performance. Additionally, this work addresses solutions for real industrial applications, being aligned with some characteristics of the Industrie 4.0 initiative, namely the distributed intelligence and self-* methods, e.g. self-adaptation, self-organization and self-configuration.

Fossil based liquid fuels, primarily used in transportation systems, are likely to be replaced with renewable resources thanks to energy transition policies. However, shifting from stable energy production (using coal, natural gas) to highly volatile renewable production will bring a number of problems as well. On the other side, tremendous developments in solar and wind power technologies encourage energy investors to maximize their contributions over the electricity grid. This highly volatile energy resources bring a strong research question to the attention: How to benefit from excess energy? Power-to-gas seems to be a strong candidate to store excess energy. Besides, power-to-hydrogen is seen as a liquid fuel for fuel cell vehicles. This work aims to analyze hydrogen as a future fuel option. Additionally, the role of energy markets, on-site production, renewable penetrations, investment and optimization problems will be also covered under this research. This research will reveal the feasibility of hydrogen as a fuel option in future smart grid.