Dynamic preshaping is important for a high speed capturing robot to catch a moving object successfully. This paper discusses dynamic preshaping for a capturing robot driven by a single wire. To control the finger posture while approaching an object, there are three essential factors, wiring, spring distribution among joints, and mass distribution in each finger link. Giving the final link posture of finger, mass distribution, and tension with respect to time, we explore how to determine both the wiring and the spring distribution. We first show how to find them under a quasi-static motion., where all dynamic effects are neglected. We further explore an approach to find them under a dynamic motion where the tension increases fast enough.
This paper presents a humanoid robot HRP-2 developed for Humanoid Robotics Project of METI (HRP for short) . HRP-2 is designed to walk on rough terrain, to prevent the possible damages to a humanoid robot's own self in the event of tipping over, and to get up again for applying it to cooperative works with a human in the open air. In this paper, the mechanical design, the electrical design, the manufacturing method, and experimental results for confirmation of designed specifications are presented.
A new method for the teleoperation of humanoid robots which integrates human operator's intention with robot's autonomy is developed. Getting hints from human conscious and subconscious motion generations, we propose the method which generates whole body motions of a humanoid robot satisfying operator's desired movement of specific points on which the operator focuses as well as the robot's balance. The method is based on control of a whole body momentum. The effectiveness of the method is experimentally confirmed by teleoperating a real humanoid robot HRP-1S to perform balanced whole body motions satisfying operator's input command.
Dividing a software component into simpler ones is an effective way to improve the manageability and reusability. However, in the case of conventional dynamically reconfigurable software architecture, the software becomes difficult to manage by dividing the components. The reason is that dividing the software components needlessly increases the number of the concurrent processes in such architecture because the software components correspond to the concurrent processes. Thus, the software components should be independent of the concurrent processes. In this paper, we present a software architecture equipped with the dynamic reconfigurability for the sequential process within a concurrent process. A multi-agent model lies at the base of the presented architecture. In our architecture, an agent, which is concurrent process, can consist of the voluntary number of the software components. And, we can modify the configuration of an agent, that is, the combination of the software components at runtime. We also describe the implementation of the real-time control software by applying the presented architecture.
This paper describes a simple method which enables control of shape and internal movements as well as migration of a herd of autonomous mobile robots in a field with various types of obstacles. In order to facilitate such collective behaviors, this paper firstly introduces an individual law of motion for each robot employing simple virtual forces considering realization on actual robots. Then, the authors propose a method where a supervisor can control the collective behaviors of the herd in realtime only by manipulating parameters and a reference position signal which are broadcast to all the robots. The behaviors by the method is analyzed theoretically, and confirmed by simulation and by experiment using Child Machines of Super Mechano Colony Experimental System.
This paper presents a method for mechanical stiffness control of the Tendon-Driven joints of manipulator. A system fundamentally mimicking a skeletal muscle system requires a non-linear elastic element to regulate stiffness of a joint. One proposes a new actuator having a non-linear elastic element, which is called ANLES. This system gives us a free-hand for designing the elastic characteristics. At first this paper describes the equations for designing the elastic characteristics of ANLES. Next the basic formula for controlling the stiffness and posture of the multi-D.O.F. joints by multiple-ANLES throughout tendons is derived. It follows the computer simulation of 3 D.O.F. joint that is manipulated by 6 tendons or 8 tendons. The performance of controlling stiffness is evaluated by depicting stiffness ellipsoid.
In this paper, we develop the dynamics-based information processing system that is a brain-like information processing based on dynamical phenomenon. A nonlinear dynamics is designed with the polynomial configuration so that it has attractors on the closed carved lines in the N dimensional space, which means memorization, reproduction and association of the time sequence data of the whole body motions. The motion transition of the humanoid robot is realized by the transition of the dynamics from one attractor to another. In this paper, we propose a design strategy of the system that generates the humanoid whole body motions using entrainment and detrainment phenomenon of the nonlinear dynamics with a motion reduction method based on the principal component analysis.
A leg-wheel robot, Chariot II, has been developed for the purpose of moving on an unknown unexplored rough terrain without using many sensors such as visual and tactile/force sensors that need accurate and complex control and heavy calculations. The robot equips with mechanically separated wheels and legs, which allows to utilize the advantages of two mechanisms. This paper presents a set up method for legs' compliance of the leg-wheel robot to decrease the pitching and rolling movement of its body when moving on unknown unexplored rough terrains. Using the proposed method, the leg's basic compliance is adjusted in proportion to the roughness of the surface when a phase of leg changes from returning phase to supporting phase. The effectiveness of this method is confirmed by simulation and experiments.
A leg-wheel robot has mechanically separated four legs and two wheels, and it performs high mobility and stability on rough terrains. In this paper, we propose gait algorithms for trot and pace gaits, and velocity limitation method, which allow continuous locomotion of a leg-wheel robot under random velocity commands. The gait algorithms which are based on the predictive event driven method determine the leg lifting timing to avoid the legs reaching the border of the work space. The velocity limitation method limits the velocity command when it exceeds the mechanical specifications of the robot. Combined the velocity limitation method with the gait algorithms, the proposed method ensures the continuity of locomotion on trot and pace gaits. It is evaluated by simulations and experiments.
A leg-wheel robot, which has mechanically separated four legs and two wheels, performs high mobility and stability on rough terrain. In this paper, we propose a basic control method for the leg-wheel robot moving on unexplored rough terrains. When moving on a rough terrain, the compliance values and the trajectories of legs are set up in proportion to the environment without using external sensors. And the step axis mechanism that we have developed realizes the stable rolling movement of the robot. The proposed method is evaluated by simulations and experiments. Using this method combined with the predictive event driven gaits, the leg-wheel robot can move on rough terrains in arbitrary directions.