As a new type of human-robot interaction (HRI), hand gesture has many advantages such as natural operation, rich expression and not subject to environmental constraints. So it is very suitable for space human-robot interaction tasks in special and harsh environment. Considering that static hand gesture is one of the main gesture expressions in human-computer interaction, so a parallel convolution neural networks (CNNs) is designed to improve the accuracy of static hand gesture recognition in the conditions of complex background and changing illumination. In addition, the method is applied to the operation of space human-robot system with hand gesture control. Various space HRI hand gestures from different subjects are evaluated and tested, and experimental results demonstrate that the proposed method outperforms the single-channel CNN methods and other popular methods with a higher accuracy.

No abstract

No abstract

SummaryThis paper discusses a technique for identifying the joint parameters of a modular robot in order to study the dynamic characteristics of the whole structure and to realise dynamic control. A method for identifying the joint parameters of the structure applying fuzzy logic combined with a genetic algorithm has been studied using a 9-DOF modular redundant robot. A Genetic Algorithm was used in the fuzzy optimisation, which helped to avoid converging to locally optimal solutions and made the results identified much more reasonable. The joint parameters of a 9-DOF modular redundant robot have been identified.

Because of the limited working hours of astronauts in the space station, the in-cabin robot has high value in the technological validation and scientific research. Based on this requirement, we proposed and designed an Astronaut Assisted Robot(AAR) working in the space station. It can float in the space station cabin, fly autonomously, and hold a fixed position and/or posture. In addition, it also possesses environmental awareness capabilities and intelligence. Thus the AAR can assist astronauts to complete some special scientific experiments or technical tests. In this paper, the system architecture and experimental equipment of the AAR are designed firstly depending on the characteristics of space microgravity environment and the requirements of assisting astronauts missions. And then, the motion principles of the AAR are analyzed and the robot's dynamic model is established by using the Newton -Euler algorithm. Since the attitude control of the robot is the basis for its free movement, the PID Neural Network(PIDNN) algorithm, which is a kind of intelligent control algorithm, is used to design the attitude controller of the AAR. Finally, the reasonability of the robot's structural design and the availability of its attitude controllers are verified through the simulation experiments.

Abstract. Flexure-based mechanisms have been widely used for scanning tunneling microscopy, nanoimprint lithography, fast servo tool system and micro/nano manipulation. In this paper, a novel planar micromanipulation stage with large rotational displacement is proposed. The designed monolithic manipulator has three degrees of freedom (DOF), i.e. two translations along the X and Y axes and one rotation around Z axis. In order to get a large workspace, the lever mechanism is adopted to magnify the stroke of the piezoelectric actuators and also the leaf beam flexure is utilized due to its large rotational scope. Different from conventional pre-tightening mechanism, a modified pre-tightening mechanism, which is less harmful to the stacked actuators, is proposed in this paper. Taking the circular flexure hinges and leaf beam flexures hinges as revolute joints, the forward kinematics and inverse kinematics models of this stage are derived. The workspace of the micromanipulator is finally obtained, which is based on the derived kinematic models.

Aiming at the application of zooming image distance measurement to mobile robot, the related key technologies are investigated in detail in this paper. Firstly, camera parameter calibration is conducted. The camera focus, optical center displacement between two foci, principal point and aberration coefficients are calculated accurately. Then, robust feature matching based on SIFT is realized by the geometrical constraint of a zooming image. Finally, the 3D reconstruction model of zooming image is established. The experimental results based on real sample images validate the practicability of the related algorithms.

No abstract