Yoshihiko Kimuro

Self-localization of mobile robots with RFID system by using support vector machine

In recent years, RFID (radio frequency identification) system has become very popular in service industries, logistics and manufacturing, as it is an inexpensive and reliable device for automatic identification. Therefore, RFID system would be useful in a problem of mobile robot self-localization, if tags are distributed in the environment, and if the robot is equipped with a RFID reader to communicate with the tags. In this paper, we propose a novel method for learning-based localization with a RFID system by using support vector machine (SVM). In order to obtain various training data for SVM learning, a number of synthesized sensor data are generated from limited amount of real sensor data. We also propose a method that enables a user to easily place tags in effective locations. In experiments with a mobile robot, the performance of the proposed method is demonstrated.

A structured environment with sensor networks for intelligent robots

This paper describes a platform of ambient intelligence for robots working in an ordinary environment for daily human life. To enable autonomous robotic activities, vision sensors and RFID tags are distributed in the environment and are connected to a network. The data from distributed sensors are integrated, and are provided to robots to support their activities. Based on the analysis of sensor functions and data requirements from the robot, we have designed and implemented a data management system which integrates the real-time data from sensors and robots. In the experiments, decision and action of the robots in a task context have been successfully achieved by using the data provided through the data management system.

Machine learning approach to self-localization of mobile robots using RFID tag

This paper proposes a method for the self-localization of a mobile robot using a passive radio frequency identification (RFID) system and support vector machines (SVMs). Using the SVM, we do not need to perform any complicated tasks for measuring the geometric position of each RFID tags to produce a look-up table as used by conventional self-localization methods. Moreover, the method works even when several malfunctioning tags are included. The performance and accuracy of the method are confirmed by our simulation test, and we conclude that the method shows almost the same performance as that of a look-up table.

Differential decreases in c-fos and aldolase C mRNA expression in the rat cerebellum after repeated administration of methamphetamine

The effects of repeated methamphetamine administration on c-fos mRNA and aldolase C (Zebrin) mRNA expression in the rat cerebellum were investigated. A single dose of methamphetamine induced c-fos mRNA expression in granule and Purkinje cells of both anterior and posterior lobes. In the posterior lobe, in particular, c-fos mRNA signals were distributed in a parasagittal organization, like Zebrin bands. Repeated methamphetamine injections reduced methamphetamine-induced c-fos mRNA signals in the anterior hemisphere and in part of the posterior vermis (lobule VII) and posterior hemisphere. Aldolase C mRNA signals in Purkinje cells decreased only in lobules where methamphetamine-induced c-fos signals were not reduced (lobules VI and IX). Therefore, differential decreases in c-fos mRNA and aldolase C mRNA expression after repeated methamphetamine administration depend upon the localization of Purkinje cells in the cerebellum. Since c-fos mRNA and aldolase C mRNA expressions are markers of excitability and the metabolic state of Purkinje cells, respectively, hypofunction of inhibitory Purkinje cells could be induced if methamphetamine is repeatedly injected. Since repeated methamphetamine administration in this experimental paradigm increased horizontal movement and the rearing activity of rats, the hemisphere of the cerebellum may be involved in development of methamphetamine-induced motor behavioral sensitization in addition to the striatum and the nucleus accumbens.

A Supervised Learning Approach to Robot Localization Using a Short-Range RFID Sensor

This work is concerned with the problem of robot localization using standard RFID tags as landmarks and an RFID reader as a landmark sensor. A main advantage of such an RFID-based localization system is the availability of landmark ID measurement, which trivially solves the data association problem. While the main drawback of an RFID system is its low spatial accuracy. The result in this paper is an improvement of the localization accuracy for a standard short-range RFID sensor. One of the main contributions is a proposal of a machine learning approach in which multiple classifiers are trained to distinguish RFID-signal features of each location. Another contribution is a design tool for tag arrangement by which the tag configuration needs not be manually designed by the user, but can be automatically recommended by the system. The effectiveness of the proposed technique is evaluated experimentally with a real mobile robot and an RFID system.

A vision system for detecting mobile robots in office environments

Our goal is to develop a new vision system for detecting mobile robots and their motion sequences in office environments. The vision system watches a sequence of images acquired by the vision sensor, as well as a sequence of robots command transmitted via a wireless network. To identify the robot, a sequence of motion is extracted for a moving object in the image sequence, and compared with the sequence of command. Simple features are used and extracted with various types of methods for detecting moving objects. This simplicity helps improve the robustness against unpredictable uncertainties inherent in the environments. In. order to improve recognition rate per frame, the vision system interacts with the robot, to learn features of the robot. A support vector machine (SVM) is used for this learning. In addition, the system is optimized reflecting characteristics of the robot and the environment.

Improvement study for measurement accuracy on wireless shoe-type measurement device to support walking rehabilitation

We have developed a shoe-type measurement device which is able to measure gait information such as step length, width and pressure distribution while daily living. We hypothesized that a walking rehabilitation system could be realized by combining shoe-type device and comprehensively display which showed analytical results for gait with real time operation. From evaluation to first trial manufacture, it was found that our system was effective to let the patients, physicians and physical therapists know quantitative gait information. However, it was also found that there were some problems such as insufficient measurement area. Then, hardware of device was redesigned to enlarge measurement area but measurement accuracy decreased. Therefore, improvement method which could decrease measurement error was proposed and the efficacy of proposition was revealed by experimental results.

Development of a shoe-type device for collecting gait information

We are developing a measurement device by means of which gait information can be obtained that cannot be obtained by existent devices. We are also evaluating the effectiveness of the device in rehabilitation programs. The final goal of our work is to realize a space where patients can exe- cute their rehabilitation routines at any time. In this paper, we report the outline of our novel measurement device. That is to say, we describe our real-time measurement technique for measuring the position and angle of the feet to obtain the gait information, as well as our method of presenting the gait in- formation.

Selecting efficient views for visualizing robot motions

Many robots are used for cleaning, patrolling, and exploring in an indoor environment. At present, however, the ability of the robots to adapt to the environment is poor, thus they often fail in the recognition of the environment when they are in unexpected situations. In order to recover the failure, a skilled person has to monitor the robot while the robot works. One of the methods to reduce the cost of monitoring is to use a remote control system. Especially it is useful to visualize a history of the robot motions, as well as regions the robot passed through. We describe a method to indicate the history information by superimposing robot trajectories onto a sequence of sensor images. The method automatically selects images to be indicated by controlling three parameters: area of the view, precision of the indication and the number of images. The method also allows users to determine the parameters manually. Moreover, we apply the method to a floor cleaning robot system, and show that swept areas are visualized efficiently.

A 4-legged mobile robot control to observe a human behavior

Describes a 4-legged walking robot that is navigated by human gesture. The navigation process is composed of two phases. The first is a human following phase. To follow a human, the robot tracks a human head by a visual tracking system. The latter phase is the recognition of human gestures. To achieve reliable and real-time recognition of the human gesture, we focus on such gestures expressed by cyclically repeated motion of the hands. We show how these gestures are recognized successfully to control the 4 legged robot.