Daisuke Katagami

Interactive classifier system for real robot learning

We describe a fast learning method for a mobile robot which acquires autonomous behaviors from interaction between a human and a robot. We develop a behavior learning method ICS (interactive classifier system) using evolutionary computation and a mobile robot is able to quickly learn rules so that a human operator can directly teach a physical robot. Also the ICS is a novel evolutionary robotics approach, using an adaptive classifier system, to environmental changes. The ICS has two major characteristics for evolutionary robotics. For one thing, it can speedup learning by means of generating initial individuals from human-robot interaction. For another, it is a kind of incremental learning method which adds new acquired rules to priori knowledge by teaching from human-robot interaction at any time.

Active teaching for an interactive learning robot

We have proposed a fast learning method that enables a mobile robot to acquire autonomous behaviors from interaction between human and robot. In this research we develop a behavior learning method ICS (interactive classifier system) using interactive evolutionary computation considering an operators teaching cost. As a result, a mobile robot is able to quickly learn rules by directly teaching from an operator. ICS is a novel evolutionary robotics approach using classifier system. In this paper, we investigate teachers physical and mental load and proposed a teaching method based on timing of instruction using ICS.

Interactive evolutionary robotics from different viewpoints of observation

In this paper, we describe influence of viewpoints of observation in an interactive evolutionary robotics system. We have been proposed a behavior learning system ICS (Interactive Classifier System) using interactive evolutionary computation. In this system, a mobile robot is able to quickly learn rules by direct teaching of a human operator. ICS is a novel evolutionary robotics approach using a classifier system. We classify teaching methods into internal observation and external one, and investigate influence of observation methods. We have experiments based on our teaching methods in two kinds of tasks. We found that teaching methods from different viewpoints of observation change teaching efficiency because of the difference between a robots recognition and an operators one in an environment.

State space self-organization based on the interaction between basis functions

In an application of reinforcement learning to real-world problems, the function approximators are usually used to approximate the value function and the policy function. It is necessary to construct the function approximator adaptively, because the value function and the policy function change along with the progress of reinforcement learning. In this work, we propose self-organizing basis network (SOBN) which is the method that constructs a function approximator using basis functions adaptively. The proposed method constructs a basis function network by connecting neighbor bases using edges. This basis function network constrains the activating region of each basis function, and the network is modified by updating the location of each basis. Using this mutual dependence, which we call the interaction between basis functions, for searching appropriate architecture of a function approximator, SOBN self-organizes the function approximator. Assuming that the method is applied to reinforcement learning, we apply the method to the function approximation problem, and evaluate approximation performance and convergence time.

Speedup of evolutionary behavior learning with crossover depending on the usage frequency of a node

For online robot behavior learning, we propose heuristics using node usage for speedup of evolutionary learning, and verify the utility experimentally. Genetic programming (GP) is an evolutionary way to acquire a program through interaction with an environment. Since behaviors of a robot are described with a program, researches on applying GP to robot behavior learning have been activated. Unfortunately, in most of the studies, the behavior learning is done off-line using simulation, not a real robot. Because convergence of GP is slow, this makes operation of a real robot quite expensive. However, since situations out of simulation easily happens in a real world, the behavior learning with a real robot (called online learning) remains very significant. Thus, in order to make online behavior learning with GP practical, we propose a crossover method for speedup of GP using node usage of a program.

Case based online training support system for ADR mediator

This paper describes an overview of an online training support system for ADR mediators. To educate good mediators, much training is necessary, which is not easy for supervisors. As supervisors have to take care of many students, they cannot spare much time for a specific student. To train students effectively, some support system is needed.This system provides an environment for online disputation. Using this system, the supervisor and students can participate in the mediation process even if they are outside of the University. Furthermore, this system stores many disputation records in the form of XML documents as a case base, and this case base is used to navigate the mediation process. During the disputation, users can retrieve old similar scenes of disputation, and they can construct proper arguments by referring to similar scenes. Furthermore, by comparing records of disputation or by analyzing them statistically, we can get the information that help to evaluate the mediation skill.

Robotic social imitation depends on self-embodiment and self-evaluation by direct teaching under multiple instructors

We have been worked about robotic social imitation in order to learn self-behavior depending on self-embodiment through interaction with multiple human. In this paper, we propose a learning method which allows to select behavioral patterns depending on self-embodiment and self-evaluation from multiple instructors by using the robot simulator Webots. We confirmed that results demonstrated that our proposal allows to improve the representative teaching data by the clustering includes two evaluation values (the distance moved forward and the impact shock for the body).

Characterized argument agent for training partner

For the resolution of disputes, Alternative Dispute Resolution (ADR) has become a popular replacement for trials. However, for mediation to work, the mediator must undergo extensive training. To help with mediator training, we have developed an online mediation support system. In this paper, we present an overview of the system and an argument agent. The argument agent participates in moot mediation as a disputant for self-training purposes. We also explain how an agents text response can be generated by retrieving from a case base situations which are likely to be met with in dispute resolution.

Teacher's load and timing of teaching based on interactive evolutionary robotics

We have proposed a fast learning method that enables a mobile robot to acquires autonomous behaviors from interaction between human and robot. In this research we develop a behavior learning method ICS (interactive classifier system) using interactive evolutionary computation considering an operators teaching cost. As a result, a mobile robot is able to quickly learn rules by directly teaching from an operator. ICS is a novel evolutionary robotics approach using classifier system. In this paper, we investigate teachers physical and mental load and proposed a teaching method based on timing of instruction using ICS.

Let's Play Catch in Words: Online Negotiation System with a Sense of Presence Based on Haptic Interaction

In recent development of online negotiation systems, the dynamic animation display has become a familiar approach for conveying nonverbal information and emotions. Yet the lack of presence caused by just looking at remains unsolved. In this study, we bring the haptic technology into the system and realize throwing a ball in a 3D virtual environment by PHANToM haptic device in order to represent making statements in turns. In this system, subjects can express their emotions by changing the ball’s color and radius. We conduct four types of observational comparison experiments to verify the effect that the haptic interaction brings about. Results from the experiments illustrate that online negotiation involving haptic interaction can increase the sense of presence, especially in text input negotiation, the difference is significant.