Masayoshi Kanoh

A Model for Generating Facial Expressions Using Virtual Emotion Based on Simple Recurrent Network

We propose an interactive facial expression model using the Simple Recurrent Network (SRN) for achieving interactions through facial expressions between robots and human beings. The proposed model counts humans in the root systemas receiversof facialexpressions to achieve a dynamic system bi-directionally affecting humans and robots. Robots typically generate only static changes in facial expression using motion files, so they seem bored, unnatural, and strange to their users. We use interactions between robots and people to diversity the inputs of robots and use emotional state transitions of robots to reduce uniformities in output facial expressions. This paper discusses a dynamic system that causes the proposed model to learn emotionalfacialexpressions basedonthoseofhumans. Next, we regard internal states generated by the proposed model as virtual emotions and show that mixed emotions can be expressed by users’ inputs from the virtual emotional space. Moreover, based on the results of a questionnaire, we see that facial expressions adopted in the virtual emotional space of the proposed model received high rates of approval from the users.

A system for converting robot 'emotion' into facial expressions

This paper presents a method that enable a domestic robot to show emotions with its facial expressions. The previous methods using built-in facial expressions were able to show only scanty face. To express faces showing various emotion, (e.g. mixed emotions and different strengths of emotions) more facial expressions are needed. We have therefore developed a system that converts emotions into robots facial expressions automatically. They are created from emotion parameters, which represent its emotions. We show that the system can generate facial expressions reasonably

Humanoid robot control based on reinforcement learning

Many existing methods of reinforcement learning have treated tasks in a discrete low dimensional state space. However, controlling humanoid robots smooth requires a continuous high-dimensional state space. In this paper, to treat the state space, we proposed an adaptive allocation method of basis functions for reinforcement learning. Up to now, grid or incremental allocation method have been proposed for allocation of basis functions. However, these methods may cause the curse of dimensionality, and fall into local minima. On the other hand, our method avoids local minima which are assessed by the trace of activity of basis functions. That is, if current state is judged to fall into a local minimum, our method eliminates a basis function which affects the state most. Moreover our method learns with a low number of basis functions because of the elimination process. To confirm the effectiveness of our method, we used a maze task to compare our method with an existing method, which has only an allocation process. Moreover, as learning of continuous high-dimensional state spaces, our method was applied to motion control of a humanoid robot. We demonstrate that our method is capable of providing better performance than the existing method.

Reinforcement learning for motion control of humanoid robots

Many existing methods of reinforcement learning have treated tasks in a discrete low dimensional state space. However, the smooth control of humanoid robots requires a continuous high-dimensional state space. In this paper, to treat the state space, we proposed an adaptive allocation method of basis functions for reinforcement learning. Grid or incremental allocation methods have previously been proposed for allocation of basis functions. However, these methods may result in the curse of dimensionality, and a fall into local minima. On the other hand, our method avoids local minima, which are assessed by the trace of activity of basis functions. That is, if the current state is determined to have fallen into a local minimum, our method eliminates a basis function, which most affects the state. Moreover our method learns with a low number of basis functions because of the elimination process. In order to confirm the effectiveness of our method, by using computer simulation, a humanoid robot learned the motion of standing up from a chair. This motion was enabled with a small number of basis functions.

Evaluating a model for generating interactive facial expressions using simple recurrent network

To improve face-to-face interaction with robots, we developed a model for generating interactive facial expressions by using a simple recurrent network (SRN). Conventional models for robot facial expression use predefined expressions, so only a limited number of expressions can be presented. This means that the expression may not match the interaction and that the person may find the expressions monotonous. Both problems can be overcome by generating expressions dynamically. We tested this model by incorporating it into a robot and comparing the expressions generated with those of a conventional model. The results demonstrated that using our model increases the diversity of face-to-face interaction with robots.

Analyzing emotional space in sensitivity communication robot Ifbot

The Ifbot robot communicates with people by considering its own emotions and theirs. Figure 1 has a front and side view of Ifbot. It has sensibility technology. This is able to detects the emotions of its interlocutor from the tone of his/her voice and the words used. Ifbot also has unique facial-expression mechanisms. Figure 2 outlines the mechanisms, which it has 10 motors and 104 LEDs. Ifbot expresses its emotions, moods and other feelings on its face by using these mechanisms in communication. We first attempt to extract characteristics of Ifbots facial expressions by mapping these to its emotional space, and then analyze its emotional space psychologically.

Haptic Communication Robot for Urgent Notification of Hearing-Impaired People

A proposed robot provides functionality inspired by the behavior of hearing dogs, which assist hearing-impaired people by alerting them to important sounds such as fire and smoke alarms. Hearing dogs use body contact for communication with hearing-impaired people, a process heavily depending on haptic communication. The proposed robot also uses physical contact for urgent notification, like a hearing dogs touching behavior. The robot can move using its wheels and then approach and bump people. To be explicit our proposal, our bumping behavior model was simplified and an experiment was conducted with hearing-impaired participants and a prototype robot. This paper reports the experimental results and discusses effectiveness of the robots bumping behavior.

Efficient joint detection considering complexity of contours

In the field of archaeology, reconstruction of earthenware imposes a heavy task on archaeologists. In order to reduce their task, we have developed a system, which can automatically reconstruct earthenware from given potsherds. The system is required to detect a pair of segments to join from potsherds correctly, by evaluation of salient values for shape of its contours. In existent method, the salient values are calculated with fixed precision. The calculation, thereby, makes the system computationally very expensive, when the number of potsherds increases. In this paper, we introduce search control, iterative deepening, into joint detection, and propose JDID (Joint Detection with Iterative Deepening), which can efficiently detect a pair of segments to join by dynamically changing its precision of salient value calculation; depending upon complexity of shape of potsherds. We have also implemented an earthenware reconstruction system and performed drastic speedup of the reconstruction.

A self-sufficiency model using urge system

The urge theory, in which emotion constitutes an autonomous system, is used to model a self-sufficiency system for communication robots. Using this model, a robot can convey its psychological and physiological conditions to the user through facial expressions, voice, behavior, and other means thereby producing psychological interaction with the user.

Acquisition of robot control rules by evolving MDDs

A method in which multi-valued decision diagrams (MDDs) are used to acquire robot action rules is proposed. Kanoh et al. have proposed a method, which uses multi-terminal binary decision diagrams (MTBDDs), to acquire robot action rules. But the variables of MTBDDs can only take values of 0 or 1; multiple variables are needed to represent a single joint angle. This increases the number of variables and the MTBDDs that represent the action rules become complex. Here, a method that uses MDDs, in which a single variable can take on multiple values, is proposed and experimental results are shown comparing MTBDDs and MDDs through simulations to acquire robot action rules.