In this paper, we consider distributed decision-making over directed time-varying multi-agent systems. We consider an adversarial bandit problem in which a group of agents chooses an option from among multiple arms to maximize the total reward. In the proposed method, each agent cooperatively searches for the optimal arm with the highest reward by a consensus-based distributed Exp3 policy. To this end, each agent exchanges the estimation of the reward of each arm and the weight for exploitation with the nearby agents on the network. To unify the explored information of arms, each agent mixes the estimation and the weight of the nearby agents with their own values by a consensus dynamics. Then, each agent updates the probability distribution of arms by combining the Hedge algorithm and the uniform search. We show that the sublinearity of a pseudo-regret can be achieved by appropriately setting the parameters of the distributed Exp3 policy.

OGIS Research Institute and Osaka Gas have developed two intelligent tutoring systems (ITSs): PCTS (Process Control Training System) and PDTS (Power Distribution Training System). This paper describes a basic concept of an ITS for plant operation based on the experience of their development. The topics include: (1) The features and structure of PCTS (i.e., text based training and model based training, a simulation model based on OOP, an intelligent authoring system). (2) What kinds of stages are needed for training systems from the view point of cognitive science (i.e., verbal learning multiple discrimination learning, rule learning, compound rule learning problem solving). (3) How to detect trainees' missing operational steps and misoperations using the perturbation method.

In the business systems design learning environment, there may be more than one solution to any given problem. For instance, the data model will be different depending on each learner's perspective. Accordingly, group learning systems are very effective in this domain. We have developed CAMELOT (Collaborative and Multimedia Environment for Learners on Teams) [18] using the Nominal Group Technique for group problem solving. In this paper, the basic framework of the collaborative learning system and the effectiveness of collaborative learning in designing the Data Model are described. By using CAMELOT, each learner learns how to analyze through case studies and how to cooperate with his or her group in problem solving. Learners come to a deeper understanding from using CAMELOT than from studying independently because they are enabled to reach better solutions through discussion, tips from other learners, and examination of one another's works.

Osaka Gas has been investigating artificial intelligence for 6 years, focusing especially on research into ICAI. We have developed a simulation based GTS (Generic Training System) in cooperation with Vanderbilt University, in Tennessee USA. This system is one of the few practical ICAI systems used for real fields, and the coursewares can be developed without any programming skill. Topics covered in this paper include the system architecture, system modules and main features. As an example, a model-based intelligent tutoring system for power distribution training systems (PDTS) is also provided. The system is in use at several places in Osaka Gas.

In the information engineering learning environment, there may be more than one solution to any given problem. We have developed CAMELOT using the Nominal Group Technique for group problem solving. This paper describes the collaborative learning system on the Internet using discussion model, the effectiveness of collaborative learning in modeling the entity-relationship diagram within the field of information engineering, and how to apply AI technologies such as rule-based reasoning and case-based reasoning to the pedagogical strategy. By using CAMELOT, each learner learns how to analyze through case studies and how to collaborate with his or her group in problem solving. As a result. We have found evidence for the effectiveness of collaborative learning, such as getting a deeper understanding by using CAMELOT than by individual learning, because they can reach better solutions through discussion, tips from other learners, examination of one another's individual solutions, and understanding alternative solutions using case-based reasoning.