Young-Seog Kim

HUMAN-LIKE GAIT GENERATION FOR BIPED ANDROID ROBOT USING MOTION CAPTURE AND ZMP MEASUREMENT SYSTEM

This article proposes a novel strategy to generate both a human-like walking pattern and a human-like zero moment point (ZMP) trajectory for a biped android robot. In general, the motion-capture technique has been widely utilized to obtain a walking pattern that is kinematically similar to the walking of a human. However, in addition to kinematic considerations, a suitable ZMP shaping technique is necessary to apply the human gait derived by motion capturing to biped robots more effectively. In previous research by the authors, a walking pattern generation strategy was developed considering the kinematics using motion capturing and Fourier fitting. However, it was found that there were differences between the calculated ZMP trajectory of the earlier research and the measured ZMP trajectory directly derived from the sensor in this research. Therefore, the differences and their factors are analyzed and a new strategy is proposed that effectively reduces the differences between them. Finally, the proposed strategy is shown to be effective for generating human-like walking pattern and ZMP trajectory for biped android robots through stick figure simulations.

WALKING PATTERN MAPPING FROM IMPERFECT MOTION CAPTURE DATA ONTO BIPED HUMANOID ROBOTS

This paper proposes an efficient walking pattern mapping algorithm from motion capture data onto biped humanoid robots. Currently, the technology known as human motion capture is widely utilized to generate various humanlike motions in many applications, including robotics. An important thing is that several difficulties are associated with motion capture data. These include a data offset issue, noise, and drift problems due to measurement errors caused by imperfect camera calibration, and marker position. If a biped robot uses motion capture data without suitable post-processes, the walking motion of the robot will differ from an actual walking motion, and the Zero Moment Point (ZMP) will be asymmetrical and noisy, leading to unstable walking. A further difficulty exists in the walking pattern mapping process due to the different joint numbers, link sizes, and weights between a human and a robot. Although walking pattern mapping is suitable after addressing the above difficulties, a slip problem between the feet and the ground can continue to cause problems. To solve these difficulties efficiently, a Fourier fitting method is proposed in this research. Improvements of walking pattern and the ZMP trajectory are confirmed using the proposed method. Furthermore, a geometric mapping method is introduced to generate walking patterns for various biped robots while maintaining a degree of similarity to humans. By applying a no-slip constraint to the feet and modifying the joint angles through inverse kinematics, the slip problem is also solved. The effectiveness of the proposed algorithm is verified through computer simulations of two different biped robots that have different sizes, weights, walking cycles, and step lengths.

DEVELOPMENT OF MOTION CAPTURE SYSTEM USING DUAL VIDEO CAMERAS FOR THE GAIT DESIGN OF A BIPED ROBOT

This paper, describes the development of a motion capture system with novel features for biped robots. In general, motion capture is effectively utilized in the field of computer animation. In the field of humanoid robotics, the number of studies attempting to design human-like gaits by using expensive optical motion capture systems is increasing. The optical motion capture systems used in these studies have involved a large number of cameras because such systems use small-sized ball markers; hence the position accuracy of the markers and the system calibration are very significant. However, since the human walking gait is a simple periodic motion rather than a complex motion, we have developed a specialized motion capture system for this study using dual video cameras and large band-type markers without high-level system calibration in order to capture the human walking gait. In addition to its lower complexity, the proposed capture method requires only a low-cost system and has high space efficiency. An image processing algorithm is also proposed for deriving the human gait data. Finally, we verify the reliability and accuracy of our system by comparing a zero moment point (ZMP) trajectory calculated by the motion captured data with a ZMP trajectory measured by foot force sensors.

ZMP Tracking Control of an Android Robot Leg on Slope-Changing Ground Using Disturbance Observer and Dual Plant Models

This paper describes a novel zero moment point (ZMP) tracking control strategy using a disturbance observer (DOB) in the presence of ground slope change for balance control of an android robot. With regard to conventional ZMP controls, many researchers have studied ZMP tracking control strategies using an inverted pendulum model on flat level ground, and they have solved a slow response problem of nonminimum phase systems by using suitable feedforward motions called walking patterns. However, the conventional methods lead to ZMP offset errors in the presence of ground slope change; it is hence necessary to quickly eliminate the ZMP offset errors to realize robust balance control. In this paper, we rapidly eliminate the ZMP offset errors through a DOB using a model inversion for robust balance control in the presence of ground slope change. In particular, a dynamic model that uses the projected center of mass (CoM) position on the ground is additionally used as an output to solve a problem that generates an unstable pole during model inversion. Finally, the proposed control strategy is verified through MATLAB simulations and experiments using a real android leg.

WHOLE-BODY MOTION GENERATION OF ANDROID ROBOT USING MOTION CAPTURE AND NONLINEAR CONSTRAINED OPTIMIZATION

This paper describes a whole-body motion generation scheme for an android robot using motion capture and an optimization method. Android robots basically require human-like motions due to their human-like appearances. However, they have various limitations on joint angle, and joint velocity as well as different numbers of joints and dimensions compared to humans. Because of these limitations and differences, one appropriate approach is to use an optimization technique for the motion capture data. Another important issue in whole-body motion generation is the gimbal lock problem, where a degree of freedom at the three-DOF shoulder disappears. Since the gimbal lock causes two DOFs at the shoulder joint diverge, a simple and effective strategy is required to avoid the divergence. Therefore, we propose a novel algorithm using nonlinear constrained optimization with special cost functions to cope with the aforementioned problems. To verify our algorithm, we chose a fast boxing motion that has a large range of motion and frequent gimbal lock situations as well as dynamic stepping motions. We then successfully obtained a suitable boxing motion very similar to captured human motion and also derived a zero moment point (ZMP) trajectory that is realizable for a given android robot model. Finally, quantitative and qualitative evaluations in terms of kinematics and dynamics are carried out for the derived android boxing motion.

Walking pattern mapping algorithm using fourier fitting and geometric approach for biped humanoid robots

Currently, the technology known as human motion capture is widely utilized to generate human-like walking patterns in humanoid robotics. An important thing is that several difficulties are associated with motion capture data. These include a data offset issue, noise, and drift problems due to measurement errors caused by imperfect camera calibration, and marker position. If a biped robot uses motion capture data without suitable post-processes, the walking motion of the robot will differ from an actual walking motion, and the Zero Moment Point (ZMP) will be asymmetrical and noisy, leading to unstable walking. A further difficulty exists in the walking pattern mapping process due to the different joint numbers, link sizes, and weights between a human and a robot. To solve these difficulties efficiently, we propose an algorithm which uses the Fourier fitting and geometric mapping method. The effectiveness of the proposed algorithm is verified through computer simulations of two different biped robots that have different sizes, weights, walking cycles, and step lengths.