Human tracking is a fundamental research issue for mobile robot, since coexistence of human and robots is expected in the near future. In this paper, we present a new method for real-time tracking of human walking around a robot using a laser range-finder. The method converts range data with r-θ coordinates to a 2D image with x-y coordinates. Then human tracking is performed using block matching between templates, i.e. appearances of human legs, and the input range data. The view-based human tracking method has the advantage of simplicity over conventional methods which extract local minima in the range data. In addition, the proposed tracking system employs a particle filter to robustly track human in case of occlusions. Experimental results using a real robot demonstrate usefulness of the proposed method.
Mobile manipulation control method for humanoid robots provides good manipulability and stability on manipulation tasks. This method leads whole body motion and locomotion, when the manipulator tip trajectory is decided. For the dexterous and powerful manipulation, the robot has to cope with the expected and unexpected forces acting to the hands. By assuming the whole body balance, we define the “ZMP Based Reference Center of Mass (ZBC) ”, which is projection of the Center of Mass (COM) to the ground of world coordinate frame, when the external forces act to the end-effectors. We propose to use the ZBC for modification control of COM with balancing control. The method is implemented to the conventional mobile manipulation control method and we construct the system which consists of four controllers: “Foothold determination”, “Walking pattern generator”, “COM modification controller” and “Momentum controller”. The experimental results show that a humanoid robot can push a heavy object by changing the foot stamps and the COM position.
In this paper, with a compass-like biped model, a memory-based controller for a planar bipedal walking system is presented. This paper is structured into two parts: Firstly, the passive dynamics of the compass-like biped is presented-to allow continuous stepping, the essential relationship between step length and initial push-off speed for ballistic motion is explored. The dynamics is enhanced and utilized via additional actuation by the controller. Secondly, the memory based walking motion generation and controller is presented, this controller refers to a memorized trajectory library, made of neutral orbits with various gait parameters for continuous walking. The proposed controller is examined and evaluated with dynamical simulations.
A passive walker with knees can walk down shallow slope in a natural gait and can exhibit a stable limit cycle only by interaction between the nonlinear dynamic system and the environment. Though the passive walker is simple, it is a sort of hybrid system which combines the continuous dynamics of leg-swing motion and the discrete event of leg-exchange. This study aims to construct the general design framework of realizing the natural and efficient walking on level ground and uphill. In this paper, we focus on the stability mechanism of fixed point in passive walking. At first, a generation method of fixed point based on its physical structure, which is formed by an energy balance, a leg-exchange phenomenon, and a leg-swing motion, is proposed. Secondly, for the purpose of highest local stability of the fixed point, a dynamics-based control method utilizing the fixed point's stability mechanism is proposed. The validity of the proposed methods is confirmed by the simulation of finite time settling in the level-walking.
The purpose of this study is to develop a wearable power assist device for hand grasping in the activity of daily living of aged or disable person. In a wearable device, a mechanism is required to have a human friendliness such as a safety, a lightweight to prevent an accident and to improve the wearable condition. In order to satisfy these requirements, extended curved and linear type pneumatic rubber artificial muscles are developed as actuators for this glove. By using the rubber muscles, the glove can assist various daily finger tasks owing to its flexibility and lightweight. In this paper, the structures of two types of rubber muscles and the power assist glove are described, and the characteristics of these devices are discussed. Finally, the effectiveness of the glove to decrease the muscular fatigue is experimentally evaluated.
We have presented safety issues such as deadlocks and dependency as well as their solutions in a dynamic updating method of running robot motion controller programs. Case studies of a mobile manipulator, a hexapod robot and a biped robot with wheeled legs illustrate discontinuity and overhead time in updating. We have also shown that the present method may significantly reduce work time of testing phase in system development.
In this paper, a trajectory tracking control method for the tendon-driven robot is presented. In the human motor control system, it is known that the spinal reflex and the cerebellum play a key role to achieve the desired limb motions and have some characteristic structures different from the conventional robot control systems, such as the internal positive feedback loops. From the viewpoint of control, the positive feedback is inconvenient since it might potentially destabilize the limb motion. However, in the biological system, the positive feedback is actively used rather than avoided and has significant role to achieve the natural human limb motions. Thus, in this paper, we attempted to establish a trajectory tracking controller for the tendon-driven robots with reference to the biological control system composed of the spinal cord and cerebellum. The stability of the controlled system was proven along the Popov's hyper-stability theory. The effectiveness of the proposed controller was demonstrated through some computer simulations and experiments using a 2 link tendon-driven robot.
In this paper, a new control method for a planar bipedal robot, which we call here Graph-based Model Predictive Control, is proposed. This method consists of two phases: the graph construction phase and the realtime control phase. In the graph construction phase, a directed graph on the state space of the control target is constructed off line. In the realtime control phase, the controller drives the state of the control target so as to make it move through graph nodes connected by directed edges. By tracing directed edges, a model predictive control is achieved in some sense. Moreover, since the directed graph is constructed in advance, the realtime computational cost is dramatically reduced compared with the ordinary MPC. In addition, by constructing multiple directed graphs based on different cost functions, one can design multiple motions and switching trajectories among them in a uniform way. The proposed method is applied to the speed changing control problem of a bipedal walker on a 2-dimensional plane and its effectiveness is verified by numerical simulation.
This paper proposes a dynamic capturing strategy where a 2D stick-shaped object with both translational and rotational velocities is completely stopped by two robotic fingers. We first show the fingertip position and the object orientation for generating a desired translational velocity or a desired rotational velocity under friction independent collision. Once the object results in a pure translational motion whose direction is perpendicular to the longitudinal axis of object, it is guaranteed that two fingers can always capture the object irrespective of friction coefficient. By using this nature, we show the friction independent capturing strategies for an object with both translational and rotational motions. The proposed strategies are demonstrated by experiments for verifying our idea.