A principled framework for general adaptive intelligent systems is described and applied to the domain of social robotics. Under the principled framework, the author develops computational methods... Show More

A principled framework for general adaptive intelligent systems is described and applied to the domain of social robotics. Under the principled framework, the author develops computational methods to address an important aspect of a social robot, which is the ability to rapidly adapt to changes in the environment such as the introduction of novel objects and installations that serve novel purposes. Methods are also developed to address another important aspect of a social robot, which is the ability to understand the needs of humans that it interacts with by having a deep model of their needs, which enables the robot to assist humans in various tasks in a socially realistic manner. The author describes the methods of causal learning and script learning through computational visual observation that allow a robot to acquire the scripts and plans that enable it to understand the intentions of humans as well as solve problems to provide assistance to humans. The robot thus adapts rapidly to changing environmental factors as new observation provides new knowledge to guide its behavior. The assistance provided to humans is formulated as a script interaction problem and the optimal points at which assistance is provided are computed using a motivational strength model derived from psychological research and formulated computationally for robotic purposes. Also, a method is proposed to handle competition of needs which arises frequently in the course of robot-human interactions to generate socially realistic and appropriate behavior on the part of the robot. This paper uses primarily a home environment to demonstrate the methodology involved, but a robot that incorporates the methodology described could rapidly adapt to any environments such as the office and factory.

Nowadays, various robot partners have been developed to realize human-friendly interactions. In general, a robot system is composed of hardware modules, software modules, and application contents.... Show More

Nowadays, various robot partners have been developed to realize human-friendly interactions. In general, a robot system is composed of hardware modules, software modules, and application contents. It takes much time to design utterance contents and motion patterns as application contents simultaneously, but the design support systems mainly focus on the generation of robot motion patterns. Furthermore, a methodology is needed to easily change the specification of hardware and software according to diversified needs, and the developmental environment to design the application contents on verbal and nonverbal communication with people. In this paper, the authors propose robot partners with the modularized architecture of hardware and software by using smart devices, and propose a developmental environment to realize easy contents design of verbal and nonverbal communication. In order to solve the problem of difficulty in the content design, they develop a design support environment using design templates of communication application contents. Next, they apply the robot partner to navigate visitors to the robot contest of the system design forum held in Tokyo Metropolitan University. Finally, they show several examples of the interaction cases, and discuss the interaction design for smart device based robot partners.

A NAO humanoid robot is programmed to act as an autonomous exercise instructor at a senior living community. In an on-site session, the robot does (i) a warm-up routine in which the robot directs... Show More

A NAO humanoid robot is programmed to act as an autonomous exercise instructor at a senior living community. In an on-site session, the robot does (i) a warm-up routine in which the robot directs participants to ask it to perform various tasks such as dancing and reciting poems and (ii) an exercise routine in which the robot guides participants through various physical exercises such as leg, hand, and neck exercises. The participants include six elderly residents, three nurses/caregivers, and two administrators. The elderly group is categorized with respect to cognitive awareness and physical capability. The session is videoed and then analyzed to measure several dimensions of human-robot interaction with these diverse participants, including affective reaction, effective reaction, and group responsiveness. Following the exercise session, a focus group session is conducted with the seniors and a separate focus group session conducted with the nurses and administrators to glean further data.

This paper proposes a cognitive architecture for building a topological map incrementally inspired by beta oscillations during place cell learning in hippocampus. The proposed architecture consists... Show More

This paper proposes a cognitive architecture for building a topological map incrementally inspired by beta oscillations during place cell learning in hippocampus. The proposed architecture consists of two layer: the short-term memory layer and the long-term memory layer. The short-term memory layer emulates the entorhinal and the ? is the orientation system; the long-term memory layer emulates the hippocampus. Nodes in the topological map represent place cells (robot location), links connect nodes and store robot action (i.e. adjacent angle between connected nodes). The proposed method is formed by multiple Gaussian Adaptive Resonance Theory to receive data from various sensors for the map building. It consists of input layer and memory layer. The input layer obtains sensor data and incrementally categorizes the acquired information as topological nodes temporarily (short-term memory). In the long-term memory layer, the categorized information will be associated with robot actions to form the topological map (long-term memory). The advantages of the proposed method are: 1) it is a cognitive model that does not require human defined information and advanced knowledge to implement in a natural environment; 2) it can generate the map by processing various sensors data simultaneously in continuous space that is important for real world implementation; and 3) it is an incremental and unsupervised learning approach. Thus, the authors combine their Topological Gaussian ARTs method (TGARTs) with fuzzy motion planning to constitute a basis for mobile robot navigation in environment with slightly changes. Finally, the proposed approach was verified with several simulations using standardized benchmark datasets and real robot implementation.

In recent years, autonomous robots become to be desired to treat multi-task. A robot must decide a concrete action for plural objectives. Major researches try to realize this by weighted rewards.... Show More

In recent years, autonomous robots become to be desired to treat multi-task. A robot must decide a concrete action for plural objectives. Major researches try to realize this by weighted rewards. Weighted rewards can represent a human's intention easily. But weight of each task must change dynamically by a change of surrounding situation or of a robot status. Authors consider an independent learning for each task and selection of one concrete action from candidates of each learning. Authors propose a priority function to calculate priority for each task corresponding to surrounding situation or a robot status and propose a system which do decision making by using the priority function. Authors confirmed the usefulness of proposed method with simulation.