Hiroki Dobashi

Robust grasping strategy for assembling parts in various shapes

In a robotic cell, assembly robots have to grasp parts in various shapes robustly and accurately even under some uncertainties in the initial poses of the parts. For this purpose, it is necessary to develop a universal robotic hand and robust grasping strategies, i.e. finger motions that can achieve planned grasping robustly against the initial pose uncertainty of parts. In this paper, we propose a methodology to plan robust grasping strategies of a universal robotic hand for assembling parts in various shapes. In our approach, parts are aligned toward planned configurations during grasping actions, and the robustness of grasping strategies is analyzed and evaluated based on pushing operation analysis. As an application example, we plan robust grasping strategies for assembling a three-dimensional puzzle, and experimentally verify the robustness and effectiveness of the planned strategies for this assembly task. Graphical Abstract

Consideration on mechanical properties of a Redundant Drive manipulator for reducing the endpoint inertia

The basic structure and the impedance control of Redundant Drive Joints with Double Actuation, RDJ-DAs, have been proposed to produce compliant joint motions with a higher bandwidth. Although RDJ-DAs could reduce the inertia of the actuators, it cannot reduce the inertia of the output link. In this paper, we propose a basic structure of Redundant Drive Manipulators, RDMs, which can reduce the total endpoint inertia. Therefore, the RDMs can be proper to produce compliant motions with a higher bandwidth. First, a problem statement on mechanical design in order to produce compliant motions is described using a basic 1-DOF manipulator, where its endpoint inertia is calculated under the specified condition to produce the endpoint force. Second, we discuss how a proposed RDM can reduce a set of joint and link inertias along the moving direction, comparing the inertias of a simple 1-DOF manipulator, a 1-DOF manipulator with RDJ-DA, and a proposed RDM. Then, the effects of the design parameter which determines the ratio of the joint torque used for driving the second joint are discussed in details.

Design of a peristaltic crawling robot using 3-D link mechanisms

In disaster areas, rescue work conducted by humans is extremely difficult. Therefore, rescue work using rescue robots in place of humans is attracting attention. This study specifically examines peristaltic crawling, the movement mechanism of an earthworm, because it can enable movement through narrow spaces and because it can provide stable movement according to various difficult environments. We developed a robot using peristalsis characteristics and derived a robot motion pattern using Q-learning, a mode of reinforcement learning. Moreover, we designed each part of the robot based on required specifications and thereby developed a real robot. We present results of motion experiments assessing the robot’s level ground movement.

Feasibility study on EEG driven robotic system to realize efficient stroke rehabilitation

We aim to develop an efficient robotic system for stroke rehabilitation, in which a robotic arm moves the hemiplegic upper limb when the patient tries to move it. In order to achieve this goal we have considered a method to detect the motion intention of the patient using EEG (Electroencephalogram), and have designed a rehabilitation robot based on a Redundant Drive Method. In this paper, we propose an EEG driven rehabilitation robot system and present initial results evaluating the feasibility of the proposed system.

Consideration on a new BCI method using EEG to detect motion intention

In our research, we aim to reconstruct the closed loop of the brain and the body for stroke rehabilitation. Stroke patients need to be supported externally when trying to move paralyzed parts of their bodies. In order to know the onset time of voluntary motions of patients, the motion intention needs to be detected. This paper considers a new BCI method using EEG to detect motion intention. The proposed method has two characteristics in EEG analysis: i) it can reduce the influence of noises; ii) it can deal with individual differences of users. The utility of the proposed method is verified with measured EEG in experiments.

Utilization of kinematical redundancy of a rehabilitation robot to produce compliant motions under limitation on actuator performance

This paper addresses the mechanical structure and control method of a redundant drive robot (RDR) to produce compliant motions, and show how the design parameters of the RDR can effect the produced motions and the mechanical and performance limitations of the actuators of the RDR. The structure and control method of the RDR can have been proper to produce compliant motions, but the effect of the design parameters of the RDR to the mechanical and performance limitations have not been clear. Therefore, the feasibility of producing compliant motions in the case of the prototype of the RDR is confirmed by conducting simulations and experiments, and then the design parameters of the RDR to the mechanical and performance limitations are verified by conducting simulations.

Orientational transition analysis toward reorientation of parts by parallel gripper allowing release motion in unstable orientations

In a versatile, flexible robotic assembly system, robots need to reorient parts from the initial orientations to another ones appropriate for assembly tasks. Previously, we have proposed a motion strategy of a parallel gripper to reorient a part allowing release motion in its unstable orientation on a flat workbench. Toward motion planning for reorientation tasks by this strategy, it is necessary to know from which orientation a part can be released so that it will settle in a desired orientation. This paper presents algorithms to obtain the relation between the released and final orientations of prismatic parts with arbitrary polygonal cross sections in a systematic way. The utility of the proposed algorithms is shown with numerical examples, and their validity is verified by experiments.

Effects of knee alignment on human gait based on wireless sensors

This study has been conducted to measure human gait locomotion according to different lower extremity alignments, and to assess the alignment effects on knees and leg muscles. Subjects were separated into three groups by knee alignment as normal knee, genu varum, and genu valgum. We conducted a synchronisation experiment using motion capture, tactile sensor and muscle potential in order to investigate the effect of the muscle activity due to differences in the alignment of the knee, and we are considering the difference in the motion analysis during walking, ground reaction and EMG. The results, were differences in the EMG and walking motion by knee types.

Acquisition of a peristaltic crawling robot's motion pattern using reinforcement learning

In disaster areas, rescue work by humans is extremely difficult. Therefore, rescue work using rescue robots in place of humans is attracting attention. This study specifically examines peristaltic crawling, the movement mechanism of an earthworm, because it can enable movement through narrow spaces and because it can provide stable movement according to various difficult environments. We develop a robot using peristalsis characteristics and derive a robot motion pattern using Q-learning, a mode of reinforcement learning. Additionally, we confirmed the convergence to the most suitable solution by coordinating Q-learning parameters.