We have been aiming at developing a robotic aid for disabled persons who need assistance in rising from a bed or a chair. In order to decrease the force applied to the body when a robotic arm contacts the human body, the mechanical behavior of the body's surface was investigated. From the experimental results, when a plate at the end of the robotic arm slides on the body's surface, shearing and normal force applied to the body were in proportion to the square of compressive displacement of the body's surface. The sliding speed showed no contribution to change in the shearing force. Furthermore utilizing this data, the most suitable joint compliance of the robotic aid touching softly against the human body was calculated. For a simulated task, the handling arm of the robotic aid was inserted between the body and a bed mat. During this process, the standards of evaluation were the normal and shearing force applied to the body. As a result, the most suitable joint compliance for the joint that was the nearest to the body was found to be from about 0.05 [rad N-1m-1].
In this paper we propose a 1DOF master-slave system called HandShake Device (HSD). HSD is composed of 2 systems: one is a master system, which is grasped by an operator, the other is a slave system, which grasps the operator's palm. Two HSDs are respectively placed at a local site and a remote site so that people can physically communicate (by shaking hands) each other through HSDs. HSD applies the virtual model method as an algorithm of tele-operation (tele-handshake). HSD was exhibited in SIGGRAPH'95 with Network Neuro Baby (NNB) which is an Artificial Life acting as a network agent in the computer network. HSD is an interface device which an operator can communicate with NNB physically.
It is important to assist doctors operating the intravascular surgical tools such as a catheter that is designed for minimum invasive surgery inside complex and narrow brain blood vessels. To solve this problem, we propose a multimedia tele-medicine for diagnosis, training, and assistance in surgery. We built a prototype of a virtual simulator system for the intravascular neurosurgery that consists of a 3D-Computer Graphics simulator and a joystick. The joystick is used for the controller and force display of catheter head direction. A visual assistance method is proposed to assist the operator. Finally, we present results of teleoperation experiments between Nagoya and Tokyo, about 350 km away each other, using high speed optical fiber network with ATM, and evaluate effectiveness of the proposed teleoperation system and the visual assistance method.
This paper proposes a method to keep stability of robotic manipulators with network connection, which are doing cooperative tasks. Such robotic manipulator systems have many possibilities, but they sometimes fail into unstable phenomenon due to the double connection: mechanical and informational connection. To aviod such unstable oscillation, the proposed method utilizes a kind of torque limiter and a notch filter.
In a teleoperation system, providing force information to a human operator can improve task performance. When a communication block between a master and a slave has transmission delay, the system is easily destabilized. Anderson and Spong guaranteed passivity in the communication block by using scattering transformation and overcame this instability caused by the time delay. But this method can be applied to the communication block with constant time delay. In traditional teleoperation system, its communication block has constant time delay. But time delay irregularly changes in a computer network because many users share telecommunication lines. This paper shows that the variable time delay destabilizes bilateral master slave manipulator with scattering transformation and a new compensation method which keeps the time delay constant. This new compensation method has been implemented in a single axis master slave manipulator.
This paper describes a visual learning method for recognizing partially occluded objects using the eigen-space analysis. Although effective in recognizing an isolated object, as was shown by Murase and Nayar, the current method can not be applied to partially occluded objects. The analysis also requires that the object is centered in an image before recognition. These limitations of the eigen-space analysis are due to the fact that the whole appearance of an object is utilized as a template for the analysis. We propose a new method, referred to as the “eigen-window” iriethod, that stores multiple partial appearances of an object in the eigen-space. Such partial appearances require a large number of memory space. To reduce the memory requirement by avoiding redundant windows and to select only effective windows to be stored, a similarity measure among windows is developed. Using a pose clustering method among windows, the method determines the pose of an object and the object type of itself. We have implemented the method and verify the validity of the method.
This paper deals with some topics of the power assist system which is used for attenuating the load force. In such system, it is a serious problem that the maneuverability and the stability are lost when the actuators are saturated. For avoiding that, we have proposed the power assist system with individual compensation ratios for gravity and dynamic load. In this paper, based on the operational sensation, we confirm the validity of the proposed method and discuss how the compensation ratios should be determined through our experiments.
Robots have played mainly important roles in factries up to the present and will be active in our daily life in future in a house. These are home robots or service robots. In order to cooperate with human being, these robots must act with fertile emotions. In order to clarify the actions with emotion, the actions of a Bunraku puppet were analyzed. Bunraku is a classical puppet entertainment in Japan. In this show, the puppets express the exaggerated human actions with fertile emotions. In this analysis, the puppet actions are modeled by stochastic processes. We present a hypothesis that “the action is decomposed into three factors: (1) the functional factor, (2) the emotional factor and (3) the stochastic factor”. We measured the real actions of a Bunraku puppet and obtained the experimental data (time series). From these time series, the above three factors were computed, based on the above hypothesis. In order to examine the validity of the above computation (signal processing), the power spectra of the stochastic factors were analyzed.
This paper proposes the circuit construction for implementing cellular-automaton LSIs. A cellular automaton is a parallel and distributed architecture suitable for high-speed image processing. To develop cellular-automaton LSIs, we must first create an unit cell circuit that can materialize cell-cell interaction rules in small-sized circuit construction. We propose constructing the cell circuits that uses νMOS FET devices. The template matching is implemented by combining multi-input νMOS circuits and inverters. We design the cell circuit for picture thinning and shrinking, and analyze its operation using a circuit simulator. A high-speed operation up to 100-MHz clock frequency can be obtained.
A modified measurement method to observe the electric earth potential difference is proposed with regard to the reduction of the unnecessary signals induced by artificial and natural earth current sources. With this method, the potential difference is measured as the voltage between two electrodes that are buried at vertically different positions below the ground surface. We examined this method at four observatories with different measurement conditions and discussed the characteristic features of the three types of the observed signals. (1) With the variation of the geomagnetic fields, the horizontal earth potential difference signals appear according to its induced current. The vertical signals, however, do not because of no induced vertical current. To be sure to realize this effect, the observation area must be sufficiently level. (2) After heavy rain fall the characteristic quasiperiodic pulse signals appear extremely rarely. No horizontal potential difference signals could be found when measured together with the vertical signals. (3) Pulse signals appear with the exactly same period that might be caused by artificial current source from structures and buildings. Consequently, to use this vertical observation method effectively, we should choose a location where the ground surface is as level as possible and as far as possible from current source points.
We have developed a highly sensitive two-dimensional (2D) polarimeter. This polarimeter was successfully used to measure the 2D birefringence distribution in thermally-induced-birefringence-compensated LD pumped Nd: glass (HAP4) laser under lasing (1054nm) and nonlasing condition. The energy loss due to birefringence was estimated quantitatively using Jones matrix. The thermal birefringence was reduced under the lasing condition compared to the non lasing condition. By inserting a 90 degree quartz rotator between two identically LD pumped Nd: glass rods, almost perfect birefringence compensation was achieved with very little residual birefringence. A considerable reduction of birefringence loss from over 12% before compensation to less than 0.4% after compensation was observed. The output performence of this laser was in a good agreement with the estimation after taking into account of the birefringence loss.
This paper clarifies, by statistic analysis, that the feature distribution extracted from on-line hand-written chinese characters is nonsymmetric and far from the gausian and also presents an adaptive Mahalanobis distance method where, under the assumption that the original distribution has one-side gausians on both sides across its average, an unknown pattern is recognized by the gausian on either side which the unknown pattern belongs to. Computational results on 160 kinds of on-line hand-written chinese character recognition are reported, which shows that the proposed method significantly improves the rate of recognition over the conventional Mahalanobis distance (7.9% for known patterns, 10.8% for unknown patterns).
This paper proposes a method to optimize automatically the structure and parameters of fuzzy (Choquet) integral models using the Bayesian framework simultaneously. It is achieved by introducing the concepts of dependency and independency of fuzzy measure in order to judge the redundancy of fuzzy measure, defining the interactive measure in fuzzy measure, and representing the Choquet integral model by interactive measure. Applying this method to numerical experiments, we proved the effectiveness of the optimization of the Choquet integral model by means of Log evidence, interactive measure and its identifying method, and the optimization of parameters by means of the significance evaluation index for interactive measure. This method enables to realize more objective representation of the ordering of the significance of fuzzy measure and more rapid optimization of fuzzy integral models which is hardly affected by errors and can increase the range of its use than conventional methods.
This paper describes the environmental learning system using a method of teaching for support of detecting errors. The system has six parts that are the system control module, the CRT image subsystem, the student learning process database, the input judgement unit, the student model, and the I/O interface. In this paper, first, the outline of the system is shown. Next, we carried out an experiment for the evaluate of the system. The purpose of the experiment is confirm the validity of a method of teaching and the conditions of the system. In the experiment, the learning system is evaluated that agreeable for learners. Finally, the system is studied based on results again.
An immune system has powerful abilities such as memory, recognition and learning to respond to invading antigens, and is expected to be applicable to many engineering applications. This paper proposes an optimization algorithm imitating the immune system to solve the multi-optimization problem partly using a genetic algorithm. The proposed algorithm is shown to be effective for searching for a set of solutions, but not local solutions, through illustrative examples of multimodal functions such as Shubert function.
The dynamics of either associative or hierarchical neural network can boil down to the discovery of fixed point or contraction to the already established fixed point in a discrete dynamical system, and strict mathematical calculations have already proven this point. In other words, the dynamics of all types of neural network can be analyzed and explained by using the fixed point theory in the traditional discrete dynamical system. What is important is that the fixed point theory also can identify the existence of chaos in the time-space and analyze its characteristics. This has provided an important link between chaos and neural network in the traditional discrete dynamical system, namely, the fixed point theory. Based on this idea, this paper proposes a method to combine chaos with neural network using the fixed point theory. With a view to practical application, the paper provides several examples on improving pattern recognition ability by adding chaotic noise in learning machines as well as on improving the ability of optimal solution in the large by creating a new Chaotic Hopfield Neural Network. The approach proposed by this paper is proved user friendly and universally applicable through lab experiments of pattern recognition and solution of the Traveling Salesman Problem on a set of 100 cities.
This paper proposes new adaptive control schemes with neural networks (NN) for Wiener type of nonlinear systems which have output nonlinearity. Firstly, by adopting a robust adaptive control law and a functional link network (FLN), we present an adaptive linearizing scheme as a primary step for a model reference control scheme, where the FLN compensates the output nonlinearity. Secondly, we analyze the stability of the adaptive linearizing scheme by using a robust adaptive control technique, and clarify that all the parameters are bounded and that the boundedness of all the signals in the closed loop is guaranteed under some reasonable conditions. Thirdly, based on the linearizing scheme, we present a new direct model reference adaptive control scheme by choosing the reference output appropriately. Moreover, the stability of the system is guaranteed under several conditions in a similar manner. Finally, we illustrate the effectiveness of the proposed scheme through some numerical examples.
In design of Client Server System (CSS), rapid progress of technology prevents an individual System Engineer (SE) from utilizing effective cases. For various combinations of hardware and software and difficult requierment analysis, it is difficult for SE to acquire CSS design technique and it is useful to support CSS design with computer. We propose a CSS configuration design support system which consists of following four parts, a requirement analysis part, a design suggestion part, a design analysis-explanation part and a learning part. Especially, the design analysis-explanation part makes educational support to SE for CSS design, because the part explains how suggested CSS satisfies requirements. This explanation is written in natural language as a scenario which shows how the suggested CSS is produced. The scenario is generated based on structural models which show the process of requirement analysis and design assumptions. To generate concise scenarios, effective information is selected using the models.
This paper proposes an algorithm of neural networks for an estimate of advertising effect. The basic premise under this algoritm is that the relationship between input and output is monotonic increase. The algorithm is useful for the domain like an estimate of advertising effect which only deficient data are available. The experimental simulation shows that the algorithm builts a reasonable model for an estimate of advertising effect.
To support the routine operations of electric power facilities, a telecommunication system designed with fail-safe features was developed using speech interface. The system was installed in the field offices of an electric power company for field testing. And we collected many speech data, HRA-ET (Human Reliability Analysis-Event Tree) is used and system reliability was evaluated. As a result, we were able to get the visibility of thing that speech interface technology is applied to the field of operation of electric power systems. This paper describes a result evaluating reliability of the speech interface technology in the field of operation of electric power systems.
Lanthanum-modified lead zirconate titanate (PLZT) thin films (50-200nm) were deposited on Pt/SiO2/Si substrate by metalorganic chemical vapor deposition (MOCVD). The electrical properties of the films were investigated as a function of the La content or the substrate temperature. Ferroelectric PZT (0/50/50) films were obtained at a substrate temperature as low as 500°C, and their electric characteristics were improved with increasing substrate temperature. La was adequately solid-dissolved into the PZT above 650°C. PLZT (15/45/55) films having a thickness of 100nm were found to have good properties for application to capacitors of dynamic random access memory, (DRAM), i, e., effective charge density of 80fF/μm2, dielectric constant of 1000, SiO2 equivalent thickness of 0.4nm and leakage current density of 5×10-8A/cm2. La addition to PZT was effective in reducing the leakage current with an increase in the registration rate. RuO2 and IrO2 bottom electrodes for ferroelectric PLZT films were also investgated. The RuO2 films were found to serve as effective diffusion barriers for NIT and MgO. Significant interdiffusion at RuO2/Si and RuO2/SiO2 interfaces occurred during the deposition of PLZT films, and annealing of RuO2 film considerably depressed the interface reactions.
Recently. ITS (Intelligent Transport Systems) project is widely developed for increasing traffic conjections and accidents. Various kinds of methods have been proposed by many researchers. In this thesis. as one of these works, we propose an image processing system mounted on a vehicle that gives infomations such as the shape of roads, ego-position of the vehicle and the distance to preceeding vehicles. We realized the system with the use of new proposed method which uses voting along the time axis to get a high degree of robustness. The system is evaluated by the experiment with real road image sequences.