Hiromitsu Hattori

Learning From Humans: Agent Modeling With Individual Human Behaviors

Multiagent-based simulation (MABS) is a very active interdisciplinary area bridging multiagent research and social science. The key technology to conduct truly useful MABS is agent modeling for reproducing realistic behaviors. In order to make agent models realistic, it seems natural to learn from human behavior in the real world. The challenge presented in this paper is to obtain an individual behavior model by using participatory modeling in the traffic domain. We show a methodology that can elicit prior knowledge for explaining human driving behavior in specific environments, and then construct a driving behavior model based on the set of prior knowledge. In the real world, human drivers often perform unintentional actions, and occasionally, they have no logical reason for their actions. In these cases, we cannot rely on prior knowledge to explain them. We are forced to construct a behavior model with an insufficient amount of knowledge to reproduce the driving behavior. To construct such individual driving behavior model, we take the approach of using knowledge from others to complement the lack of knowledge from the target. To clarify that the behavior model including prior knowledge from others offers individuality in driving behavior, we experimentally confirm that the driving behaviors reproduced by the hybrid model correlate reasonably well with human behavior.

Advanced Agent Technology : Aamas Workshops 2011, Ample, Aose, Arms, Docmas, Itmas, Taipei, Taiwan, May 2-6, 2011. Revised Selected Papers

Advanced Agent Technology : Aamas Workshops 2011, Ample, Aose, Arms, Docmas, Itmas, Taipei, Taiwan, May 2-6, 2011. Revised Selected Papers

A Multi-Phase Protocol for Negotiation with Interdependent Issues

Multi-issue negotiations are a central component of many important coordination challenges. Almost all previous work in this area has assumed that negotiation issues are independent, making it relatively easy to find high-quality agreements. In many real-world problem domains, however, issues are interdependent, making hard to find good agreement due to the nonlinearity of the agents utility functions. The key challenge, in this context, is finding high-quality agreements without making unrealistic demands concerning how much agents reveal about their utilities. In this paper, we propose a protocol wherein the negotiating agents, working with the mediator, progress through a multi-phase narrowing of the space of possible agreements. We show that our method outperforms existing methods in large nonlinear utility spaces.

Participatory technologies for designing ambient intelligence systems

To realize large-scale socially embedded ambient intelligence systems, this paper proposes a design methodology towards society-centered design. Participatory technologies and multiagent systems are essential in the new system design perspective. Multiagent systems make it possible to test and predict the behavior of socially embedded systems. We have already developed the scenario description language, which describes interaction protocols that link agents to society. We use the virtual space, wherein agents behave under given scenarios, in explaining each step of society-centered design. The process consists of participatory simulation, where agents and human-controlled avatars coexist in virtual space to jointly perform simulations, and augmented experiment, where an experiment is performed in real space by human subjects, scenario-controlled agents, and human extras. For realizing realistic interactions between agents and humans during participatory simulations, an agent model that can reproduce human-like agent behaviors is needed. We show a direction for agent modeling based on learning from humans in actual application environments.

Massive Multiagent-Based Urban Traffic Simulation with Fine-Grained Behavior Models

As it is getting easier to obtain reams of data on human behavior via ubiquitous devices, it is becoming obvious that we must work on two conflicting research directions for realizing multiagent-based social simulations; creating large-scale simulations and elaborating fine-scale human behavior models. The challenge in this paper is to achieve massively urban traffic simulations with fine-grained levels of driving behavior. Toward our objective, we show the design and implementation of a multiagent-based simulation platform, that enables us to execute massive but sophisticated multiagent traffic simulations. We show the capability of the developed platform to reproduce the urban traffic with a social experiment scenario. We investigate its potential to analyze the traffic from both macroscopic and microscopic viewpoints.

Multi-model based simulation platform for urban traffic simulation

Multiagent-based simulations are regarded as a useful technology for analyzing complex social systems; for example, traffic in a city. Traffic in a city has various aspects such as route planning on the road network and driving operations on a certain road. Both types of human behavior are being studied separately by specialists in their respective domains. We believe that traffic simulation platforms should integrate the various paradigms underlying agent decision making and the target environment. We focus on urban traffic as the target problem and attempt to realize a multiagent simulation platform based on the multi-model approach. While traffic flow simulations using simple agents are popular in the traffic domain, it has been recognized that driving behavior simulations with sophisticated agents are also beneficial. However, there is no software platform that can integrate traffic simulators dealing with different aspects of urban traffic. In this paper, we propose a traffic simulation platform that can execute citywide traffic simulations that take account of the aspects of route selection on a road network and driving behavior on individual roads. The proposed simulation platform enables the multiple aspects of city traffic to be reproduced while still retaining scalability.

A Driver Modeling Methodology Using Hypothetical Reasoning for Multiagent Traffic Simulation

We propose how to acquire drivers individual operation models using the three-dimensional driving simulator in order to implement distinct personalities on each agent. In this paper, operation models are defined as sets of prioritized operation rules, each of which consists of the world as observed by a driver and his/her next operation according to the observation. Each driver might have different set of rules and their priorities. We apply a method to acquire individual operation models using hypothetical reasoning. Because of the method, we are able to obtain operation models which can explain drivers operation during driving simulation. We show some operation models acquired from aged/young human drivers, and then clarify the proposed method can catch each drivers characteristics.

Special Issue on Intelligent Agents in Traffic and Transportation

Reference EPFL-ARTICLE-208649doi:10.1080/15472450.2013.856719View record in Web of Science Record created on 2015-05-29, modified on 2017-05-12

Participatory Simulation Environment gumonji/Q: A Network Game Empowered by Agents

Network games are attracting attention as simulation platforms for social experiments because of their rich visualization performance and scalability. Our objective in this study is to develop a participatory simulation platform on a network game. Unlike non player characters (NPCs) in network games, agents in a participatory multiagent-based simulation (PMAS) should behave as real-world humans according to behavior models. We developed a novel networked participatory simulation platform called gumonji/Q by integrating scenario description language Q with the network game gumonji. This paper details the implementation of gumonji/Q. In order to connect Q and gumonji, we implement communication sub-components that realize TCP/IP communication between them, and a scenario translator to convert a request from Q into a sequence of operators. This makes it possible for the gumonji simulator to deal with human-controlled avatars and Q-controlled agents in a unified way.

A Multi-Issue Negotiation Protocol Among Nonlinear Utility Agents: A Preliminary Report

1 Department of Computer Science and Engineering, Graduate School of Engineering, Nagoya Institute of Technology, Gokiso, Showa-ku, Nagoya 466-8555, Japan ito.takayuki@nitech.ac.jp 2 Center for Collective Intelligence, Sloan School of Management, Massachusetts Institute of Technology, Three Cambridge Center, NE20-336, Cambridge, MA 02142, USA m klein@mit.edu 3 Department of Social Informatics, Graduate School of Informatics, Kyoto University, Yoshida-Honmachi Sakyo-ku, Kyoto 606-8501, Japan