Hiroaki Ochi

Set-point control of a musculoskeletal arm by the complementary combination of a feedforward and feedback manner

This paper proposes a novel set-point control method of a musculoskeletal system by combining a feedforward and feedback manner to complement each drawback each other. In our previous work, a feedforward positioning method of the musculoskeletal arm model was proposed which does not need any realtime sensory information. Its performance, however, depends on a muscular arrangement and an attitude of the arm, and thereby a large initial muscular internal force is necessary to make a good performance. On the other hand, it is well-known that a visual servoing is effective and versatile for the set-point control. However, there is a considerable time-delay due to a computational burden to acquire useful information from an image and an insufficient sampling period to capture each image when using a video frame rate camera. Thus in this paper, the feedforward and feedback signal are linearly combined into one in order to mutually complement each drawback. The combined control signal is newly designed and then numerical simulation results are shown to demonstrate the effectiveness and usefulness of the proposed method.

Curved surface fitting method using a raster-scanning window and its application to stereolithography-based reverse engineering

In this paper, a curved surface fitting method using a raster-scanning window is proposed to smooth the original organized point cloud data (PCD) and extract the original features in the presence of noise. The method allows PCD to be smoothed while keeping its own original features. Then, a Stereolithography (STL) generator is proposed to produce triangulated patches from smoothed PCD, and this helps to deal with a triangular representation of a 3D surface geometry. The process allows reconstructing 3D digital data of a real object described using STL format from original PCD in the presence of noise for reverse engineering. The effectiveness and usefulness of the proposed techniques are demonstrated through actual experiments to smooth PCD, reconstruct STL, generate cutter location source (CLS) data and perform machining of the object model described in the original PCD.

Sensorless point-to-point control for a musculoskeletal tendon-driven manipulator: analysis of a two-DOF planar system with six tendons

Tendon-driven robot utilizes only tensile force (i.e. tension) for motion generation. Therefore, a redundant actuation is characteristically necessary, and then it yields the internal force among tendons. Given the internal force for balance at a desired posture, the musculoskeletal tendon-driven manipulator has the inherent possibility of point-to-point position control without any sensory feedback. However, the motion convergence is strongly governed by the arrangement of tendons.This study analyzes the mathematical conditions of convergence for this sensorless position control by use of a Lyapunov function. Subsequently, targeting the two-link musculoskeletal structure with six tendons, the sufficient conditions for the convergence at desired posture are further defined by employing an approximation of the tendon-length based on a Taylor expansion. Finally, the convergent conditions are verified through simulation and validated via experimental results. Graphical Abstract

A proposal of a SMA actuated wing mechanism using flexible structure for the capability of various flow speeds

This paper proposed a concept of flexible wing-mechanism for the underwater fishlike robots. The proposed system consists of a fishlike body and a pair of flexible wing. The flexible wing mechanism utilizes a Shape Memory Alloy (SMA) actuator to realize small and lightweight system with high output/weight ratio. The SMA actuator can be controlled to adjust the shape. In the flow of water, shape of the wings is controlled and then the fluid resistance will be changed. Then, we can control the posture of the robot. The wing mechanism is constructed mainly from flexible materials. Thus, range of the fluid resistance becomes different between the high speed flow and low speed flow. In the fast flow condition, flexible wing mechanism is compressed and the range of the shape control becomes low. Then, the mechanism can control the posture of the robot under the various flow speeds.

Feed-forward positioning of musculoskeletal-like robotic systems with muscular viscosity: Determination of an adequate internal force

This paper proposes a new feed-forward positioning method for a musculoskeletal-like robotic system considering a muscle-like nonlinear viscosity, and a new determination method of the internal force using the reinforcement learning scheme. In our previous works, a feed-forward positioning method for the musculoskeletal-like robotic systems has been proposed. In the method, the position regulation of the system can be accomplished by inputting a desired internal force balancing at a desired position. It has been quite effective for the muscle-like driven mechanism because no sensor is necessary to regulate the position. However, this method often induces an overshoot phenomenon when performing a set-point control. In addition, there is another intrinsic problem that musculoskeletal-like redundant-driven mechanisms own the ill-posed problems that the internal force is unable to determine uniquely. In this paper, for the farmer problem, a muscle-like nonlinear viscosity is newly added to the controller to reduce such an overshoot phenomenon and then to expand the stable region of the manipulator. For the latter problem, a determination method of the internal force using a reinforcement learning scheme is newly proposed. In what follows, firstly a new feed-forward controller which considers the muscle-like viscosity is introduced, and shows its effectiveness through numerical simulations. Next, the determination method of the internal force using a reinforcement learning scheme is proposed and its effectiveness is also shown through numerical simulations.

Study of human motion generation based on redundancy of musculoskeletal structure: Analysis of potential generated by internal force for two-link system

The human body has a musculoskeletal system with the muscles which exist around the bones and joints. Taking notice of the structural characteristics that a human possesses inherently, this paper analyzes feedforward position control for the musculoskeletal system. The feedforward positioning does not need any sensory feedback by use of internal force balancing at a desired posture. Targeting a non-pulley musculoskeletal system with two links and six muscles, this paper clarifies mathematical conditions of the feedforward positioning to converge at a desired posture. In the analysis, muscular length is approximated by Taylor expansion. Based on quasi-statical approach, the convergent conditions are clarified. The verification of the conditions is conducted through simulation.

Design Tool of Deep Convolutional Neural Network for Visual Inspection.

In this paper, a design tool for deep convolutional neural network (DCNN) is considered and developed. As a test trial, a DCNN designed by using the tool is applied to visual inspection system of resin molded articles. The defects to be inspected are crack, burr, protrusion and chipping phenomena that occur in the manufacturing process of resin molded articles. An image generator is also developed to systematically generate many similar images for training. Similar images are easily produced by rotating, translating, scaling and transforming an original image. The designed DCNN is trained using the produced images and is evaluated through classification experiments. The usefulness of the proposed design tool has been confirmed through the test trial.

Iterative curved surface fitting algorithm using a raster-scanning window

In this paper, an iterative curved surface fitting method using a small sliding window is first proposed to smooth the original organized point cloud data (PCD) with noise and fluctuation. Samples included in a small sliding window positioned in PCD are successively fitted to a quadratic surface from upper left to lower right using a least squares method. In the iterative process, outliers of samples are asymptotically removed based on an evaluation index. This proposed method allows original PCD to be smoothed keeping its own shape feature. Then, the already developed stereolithography (STL) generator is used to produce triangulated patches from the smoothed PCD. The process allows to reconstruct 3D digital data of a real object written with STL format for reverse engineering from original PCD with noise. The effectiveness and usefulness of the proposed curved surface fitting method are demonstrated through actual smoothing experiments.

Complementary compound set-point control by combining muscular internal force feedforward control and sensory feedback control including a time delay

Abstract This paper proposes a new set-point control method for a musculoskeletal arm by combining muscular internal force feedforward control with feedback control including a large time delay. The proposed method accomplishes robust and rapid positioning with a relatively small muscular force. In the positioning by the muscular internal force feedforward controller, a large muscular force is required to achieve good performance. On the other hand, in the positioning by the feedback controller including the large time delay, the system can easily fall into an unstable state. A simple linear combination of these two controllers makes it possible to improve the control performance and to overcome the drawbacks of each controller in a complementary manner. First, a two-link six-muscle arm model is considered as a musculoskeletal system in this study. Second, the new set-point control method, which consists of the feedforward control signal and the feedback control signal including the time delay, is designed. Third, the stability of the proposed method is investigated using the Lyapunov–Razumikhin method. Finally, the results of numerical simulations and experiments are presented to demonstrate the advantages of the proposed method.

Stiffness evaluation of a tendon-driven robot with variable joint stiffness mechanisms

This paper proposes a new tendon-driven robot with variable joint stiffness mechanisms. The tendon-driven robot is able to vary the stiffness of joints by sliding variable stiffness mechanisms over the link by wire tensions. As a reason for that structure and moment arms of the tendon-driven robot are changed depending on the position of the variable mechanism. Thus in this paper, the tendon-driven robot with variable stiffness mechanisms is designed, and the stiffness of the tendon-driven robot is evaluated by using a stiffness ellipsoid.