Takayuki Tanaka

Muscle stiffness sensor to control an assistance device for the disabled

An innovative sensor has been developed as a man–machine interface to control an assistance device for people with disabilities. This sensor is applied to a mechanical glove that has a grasping function, and its usefulness is demonstrated through experiments. The sensor is attached to the operator’s skin and detects noninvasively how much the muscle under the sensor has stiffened.

Development and Motion Control of a Wearable Human Assisting Robot for Nursing Use

The objective of this study is to perform certain nursing related motions by the human-HARO system that are too heavy for a nurse. This paper will present the motions designed for Human-Assisting Robot (HARO), the control algorithms developed in this study for improving the maneuverability of HARO, and the experimental results showing that HARO successfully maneuver a real patient. The safety and reliability of HARO was enhanced by significantly improving its maneuverability in such ways as limiting motor velocities, determining control parameters from mechanical characteristics, and filtering interfering forces detected by force sensors in HARO. As a result, the human-HARO system successfully performed a series of designed motions on a real patient including lifting, holding, pulling and pushing. The feasibility of using HARO for nursing care assistance is demonstrated through extensive experiments.

Soft computing algorithms for intelligent control of a mobile robot for service use

Abstract The arrangement principles and design methodology on soft computing for complex control framework of AI control system in Part 1 of this paper are developed. The basis of this methodology is computer simulation of dynamics for mechanical robotic system with the help of qualitative physics and search for possible solutions by genetic algorithms (GA). In Part 2 optimal solutions for navigation with avoidance of obstacles and technological operations as opening of door with a manipulator on GA and fuzzy neural network (FNN) are obtained and knowledge base (KB) for fuzzy controller is formed. Fuzzy qualitative simulation, GA and hierarchical node map (HN), and FNN have demonstrated their effectiveness for path planning of a mobile robot for service use. The results of fuzzy robot control simulation, monitoring, and experimental investigations are described.

Intelligence computing for direct human-robot communication using natural language and cognitive graphics

A direct human-robot communication system based on natural language (NL) and cognitive graphics (CG) as a part of a new hierarchical structure of an AI control system of a mobile robot for service use (MRSU) is developed. The software of the simulation system for direct human-robot communication and MRSU behavior based on NL and CG is described. The NL and CG are used for description and representation of possible external worlds in the robot artificial life. This system allows us to evaluate the control algorithms of real time robot behavior and to reduce difficulties connected with such troubles as robot collisions with real objects and robot hardware damage. The mathematical background of direct NL human-robot communication and robot behavior simulation system is knowledge engineering based on spatio-temporal and action logics, default reasoning, cognitive graphics and soft computing. The main concepts, structure, conceptual model of the simulation system for description of the artificial life of the MRSU are discussed. A simulation example and real experimental results are described.

Toward New Humanoid Applications: Wearable Device Evaluation Through Human Motion Reproduction

This chapter addresses a new application of humanoid robot for evaluation of wearable assistive devices. One of the issues for diffusion of assistive devices that have been developed recently is its objective and qualitative evaluation to validate its assistive effects. Human subject experiments that have frequently been used have several drawbacks such as heavy ethical procedures, repeatability and subjectivity coming from tests with questionnaires. A humanoid robot with human-like structure and shape has potential to be used as an “active mannequin.” By reproducing human motions based on a technique called retargeting, a humanoid robot can test wearable devices instead of a human. It has advantages as it can repeat the same motions to test the product under the same conditions, and it has no ethical risks. The largest advantage is its capacity of quantitative evaluation by measuring joint torques, which allows direct validation of supportive torque generated by the device. By taking an example of a supportive wear “Smart Suit Lite” supporting the user’s lower back by elastic bands, we will overview the evaluation framework of humanoid-based assistive device evaluation, including human motion retargeting, experimental device evaluation with humanoid HRP-4C, and validation of the accuracy of the results using an identification method.

Wearable Sensor System for Lumbosacral Load Estimation by Considering the Effect of External Load

Anteflexion of the spine is a crucial motion in performing many tasks during work and daily life. It is particularly important in tasks such as providing care to others and carrying objects. To devise measures for preventing back pain, it is necessary to determine the postures associated with high risk of low back pain. Postures that increase lumbosacral load should be identified to reduce the risk of low back pain. In previous work, the relationship between posture and intervertebral loading was clarified, and the centers of gravity in the upper body and the waist shape were estimated. Moreover, individual differences were considered to improve the accuracy of the estimation. This method can estimate the lumbosacral load with sufficient accuracy. However, lumbosacral loading was examined in relation to posture and increases with external load. Therefore, the external load should be included in lumbosacral load estimation. In this study, we developed a back muscle exertion estimation method by using stiffness sensors to measure back muscle exertion, because the back muscle exertion changes with the external load. We conducted experiments in which participants wore the sensor system and the lumbosacral load was estimated from the external load. Estimation using the muscle stiffness sensors was better than previous estimation methods.