Skip to main content

Über dieses Buch

This book contains revised selected and invited papers presented at the International Workshop on Massively Multi-Agent Systems, MMAS 2018, held in Stockholm, Sweden, in July 2018.
The 7 revised full papers presented were carefully reviewed and selected for inclusion in this volume. Also included are 3 post-workshop papers. The papers discuss enabling technologies, new architectures, promising applications, and challenges of massively multi-agent systems in the era of IoT. They are organized in the following topical sections: multi-agent systems and Internet of Things; architectures for massively multi-agent systems; and applications of massively multi-agent systems.



Multi-agent Systems and Internet of Things


Distributed Speaking Objects: A Case for Massive Multiagent Systems

Smart sensors and actuators, embedding learning and reasoning features and associated to everyday objects and locations, will soon densely populate our everyday environments. Being capable of understanding, reasoning, and reporting about what is happening (for sensors) and about what they can make possibly happen (for actuators), these “speaking objects” will thus be assimilable to autonomous situated agents. Accordingly, populations of speaking objects will define dense and massive multiagent systems, devoted to monitor and control our environments, let them be homes, industries or, in the large-scale, whole cities. In this context, the necessary coordination among speaking objects will be likely to become associated with the capability of argumenting about situations and about the current state of the affairs, triggering and directing proper distributed conversations, and eventually collectively reach future desirable state of the affairs. In this article, we detail the speaking objects vision, overview the key enabling technologies, and analyze the key challenges for engineering large-scale collectives of speaking objects and their conversations.
Marco Lippi, Marco Mamei, Stefano Mariani, Franco Zambonelli

Injecting (Micro)Intelligence in the IoT: Logic-Based Approaches for (M)MAS

Pervasiveness of ICT resources along with the promise of ubiquitous intelligence is pushing hard both our demand and our fears of AI: demand mandates for the ability to inject (micro) intelligence ubiquitously, fears compel the behaviour of intelligent systems to be observable, explainable, and accountable. Whereas the first wave of the new “AI Era” was mostly heralded by sub-symbolic approaches, features like explainability are better provided by symbolic techniques. In this paper we focus on logic-based approaches, and discuss their potential in pervasive scenarios like the IoT and open (M)MAS along with our latest results in the field.
Andrea Omicini, Roberta Calegari

Integrating Internet of Services and Internet of Things from a Multiagent Perspective

To realize the Internet-based sociotechnical systems, it is necessary to build a comprehensive and effective infrastructure to support the interaction between various cloud services on the Internet and the physical world in which we live. For example, the information produced by the sensors is usually aggregated, processed and analyzed by services in the cloud, which can be used by various stakeholders for decision-making in many different application fields. Therefore, we need to consider integrating the Internet of Services (IoS), which enables the flexible sharing and composition of cloud services on the Internet, with the Internet of Things (IoT), which represents the constellation of things equipped with various sensors and actuators. The integration of IoS and IoT often involves multiple parties and so must deal with complex issues such as interaction, dynamics, scalability and decision making, all of which can be studied from a multiagent perspective. In this paper, we start by discussing the necessities and challenges for integrating IoS and IoT. Then, we propose an integrated architecture and examine it from two multiagent perspectives. One is to regard the integrated architecture of IoS and IoT as a multiagent-based architecture considering various patterns of service composition and interaction. The other is to apply multiagent methodologies when designing sociotechnical systems for various application domains based on the integrated IoS/IoT architecture. Moreover, we use the example application of designing multilingual environments to discuss the above two perspectives with possible future research directions.
Donghui Lin, Yohei Murakami, Toru Ishida

Architectures for Massively Multi-agent Systems


Two-Layer Architecture for Distributed Massively Multi-agent Systems

Existing massively multi-agent systems are aimed at handling tens of thousands of agents on a single server or a computer cluster. To this end, the agents are implemented as a data structure on the server to run at high speed. However, in future IoS/IoT environments, it will be necessary to deploy agents to distributed servers. Therefore, we propose a two-layered architecture consisting of macro-agents and micro-agents: the former controls the distributed environment and the latter solves the problem cooperatively. The macro-agents pre-installed on servers form a self-organized network by communicating with neighbor macro-agents. On the other hand, micro-agents are implemented as data structures on the server and solve problems under control of the macro-agents. An example scenario is presented to illustrate how to apply the proposed architecture to driving assistance with environment-embedded sensors.
Yohei Murakami, Takao Nakaguchi, Donghui Lin, Toru Ishida

Multi-agent Social Simulation for Social Service Design

Multi-agent social simulation (MASS) can be a powerful tool for designing social systems and services. Due to increases in computational power and progress in the social big data field, we can now apply MASS to real social systems, such as urban traffic and disaster response scenarios. Here, we demonstrate several MASS applications and discuss future possibilities and issues in this emerging domain.
Itsuki Noda

Inverse Reinforcement Learning for Agents Behavior in a Crowd Simulator

Crowd behavior has been subject of study due to its applications in fields like disaster evacuation, smart town planning and business strategic placing. However, obtaining patterns from the crowd to make a working model is difficult, as it requires an enormous quantity of data from observation and analysis and is impractical in many scenarios due to logistic and legal issues. Machine learning techniques are a good tool to overcome these difficulties, using a relatively small training data set to identify patterns, allowing crowd agents to react to similar situations accordingly. We implemented a behavioral agent model that uses such techniques into a large-scale crowd simulator, and apply inverse reinforcement learning to adjust agents’ behaviors by examples. The goal of the system is to provide to the agents a realistic behavior model and a method to orient themselves without knowing the scenario’s layout, based in learnt patterns around environment features.
Nahum Alvarez, Itsuki Noda

FARM: Architecture for Distributed Agent-Based Social Simulations

In many domains, high-resolution agent-based simulations require experiments with a large number (tens or hundreds of millions) of computationally complex agents. Such large-scale experiments are usually run for efficiency on high-performance computers or clusters, and therefore agent-based simulation frameworks must support parallel distributed computations. The development of experiments with a large number of interconnected agents and a shared environment running in parallel on multiple compute nodes is especially challenging because it introduces the overhead of cross-process communications.
In this paper we discuss the parallel distributed architecture of the farm agent-based simulation framework for social network simulations. To address the issue of shared environment synchronization we used a hybrid approach that distributes the simulation environment across compute nodes and keeps the shared portions of the environment synchronized via centralized memory storage. To minimize cross-process communication overhead, we allocate agents to processes via a graph partitioning algorithm that minimizes edge cuts in the communication graph, estimated in our domain by empirical data of past agent activities. The implementation of the toolkit used off the shelf components to support centralized storage and messaging/notification services.
This architecture was used in a large-scale Github simulation with up to ten million agents. In experiments in this domain, the graph partitioning algorithm cut overall runtime by 67% on average.
Jim Blythe, Alexey Tregubov

Applications of Massively Multi-agent Systems


Diversity in Massively Multi-agent Systems: Concepts, Implementations, and Normal Accidents

Coordination for Transportation as a Service (TaaS) can be implemented on a spectrum, ranging from independent agents communicating exclusively through market exchanges to hybrid market/hierarchy approaches fixed hierarchical control systems. An overview of each approach is described and a detailed description of recent work in simulating a hybrid solution is presented. The use of diversity as a potential approach to reduce the impact of catastrophic Normal Accidents is discussed.
Philip Feldman, Antonio Bucchiarone

CARAVAN: A Framework for Comprehensive Simulations on Massive Parallel Machines

We present a software framework called CARAVAN, which was developed for comprehensive simulations on massive parallel computers. The framework runs user-developed simulators with various input parameters in parallel without requiring the knowledge of parallel programming. The framework is useful for exploring high-dimensional parameter spaces, for which sampling points must be dynamically determined based on the previous results. Possible use cases include optimization, data assimilation, and Markov-chain Monte Carlo sampling in parameter spaces. As a demonstration, we applied CARAVAN to an evacuation planning problem in an urban area. We formulated the problem as a multi-objective optimization problem, and searched for solutions using multi-agent simulations and a multi-objective evolutionary algorithm, which were developed as modules of the framework.
Yohsuke Murase, Hiroyasu Matsushima, Itsuki Noda, Tomio Kamada

BASIC: Towards a Blockchained Agent-Based SImulator for Cities

Autonomous Vehicles (AVs), drones and robots will revolutionize our way of travelling and understanding urban space. In order to operate, all of these devices are expected to collect and analyze a lot of sensitive data about our daily activities. However, current operational models for these devices have extensively relied on centralized models of managing these data. The security of these models unveiled significant issues. This paper proposes BASIC, the Blockchained Agent-based Simulator for Cities. This tool aims to verify the feasibility of the use of blockchain in simulated urban scenarios by considering the communication between agents through smart contracts. In order to test the proposed tool, we implemented a car-sharing model within the city of Cambridge (Massachusetts, USA). In this research, the relevant literature was explored, new methods were developed and different solutions were designed and tested. Finally, conclusions about the feasibility of the combination between blockchain technology and agent-based simulations were drawn.
Luana Marrocco, Eduardo Castelló Ferrer, Antonio Bucchiarone, Arnaud Grignard, Luis Alonso, Kent Larson, Alex ‘Sandy’ Pentland


Weitere Informationen

Premium Partner