Takashi Nishiyama

Development of agent system based on decision model for creating an ambient space

This paper describes a decision model for an autonomous agent that provides an inhabitant with comfort based on information network technologies that connect home electric appliances with household equipment. The inhabitant enjoys the benefit of comfort, while he pays the cost for keeping that comfort. The autonomous agent should decide and control household equipment considering that cost from the inhabitant’s viewpoint. Thus, we utilized a representation scheme called an “influence diagram” that enabled us to model the decision-making process of the agent from the inhabitant’s point of view. First, decision modeling using the influence diagram is presented via an example. The presented model consists of three information-processing modules: a module for estimating the situation of an inhabitant based on information from home networks, a module for evaluating comfort of the inhabitant, and a module for making decisions that maximize the utility of the inhabitant from both the viewpoints of comfort and the cost paid for that comfort. Next, an experiment for verifying whether the presented model is effective or not, and its results are described. Finally, our model of the agent is discussed in relation to social intelligence design by investigating the interactive processes between the agent and the inhabitant.

A decision-making modeling for home ambient intelligent system and its extension to multi-agent modeling

This paper describes a decision model for an autonomous agent that provides an inhabitant with comfort using home information networks that connect electric appliances with household equipment. The agent should manage and control electric appliances or household equipments based on information obtained from home networks considering cost paid for keeping the comfort from the inhabitants viewpoint. We utilize a representation scheme called influence diagrams that enable us to model the decision-making process of the agent. This agent modeling, however, is restricted to a single agent for a single inhabitant. We are now extending the modeling to a multi-agent modeling, which is described in the latter part of this paper.

Development of humanoid agent system employing the knowledge on transactional analysis

A humanoid agent has been studied that is employing artificial intelligence and emotions to communicate with users not only with speech but also with gestures. In order for the agent to interact with users in a smooth way, the idea of transactional analysis is introduced from the field of psychology into the agent. This work briefly describes the introduction of transactional analysis, followed by the architecture of humanoid agent that has been proposed.

Multiagent evacuation planning by distributed constraint satisfaction

A multiagent planning method is proposed, in which the agents intentions are identified in the individual plans, and the plans are coordinated to be integrated into a consistent total plan under consideration of the intention. The intention identification process is modeled as a distributed constraint satisfaction process. The model is represented as a predicate transition-net, a kind of high level Petri net. The proposed method is applied to an evacuation planning problem from a building under an emergent fire situation. The coordination process of the individual plans is implemented by the parallel processing language Occam. The gradual and event-driven manner of the process yields the flexibility and transparency of the planning method.<<ETX>>