A state-of-the-art biped robot takes footsteps that its heel or tiptoe overhangs step corner edges while ascending and descending stairs, as humans naturally do. Taking the overhang step strategy requires a real-time sensing method for estimating step edge positions in high precision. In this paper we propose a step edge estimation method using stereo images for the biped robot taking the overhang step strategy. After extracting planar areas from a 3D occupancy grid map generated from range sensor measurements, we estimate the plane parameters of each area by a fast direct method using stereo images. Then we obtain the line edge of each step so that the line segments the view area into two image regions both of which have small absolute difference values between stereo images. Our real-time estimation method enables the biped robot to smoothly walk on stairs and bumps. The validity of our method is demonstrated through on-line experiments using stereo cameras mounted on the body of a biped robot traversing real stairs.
In this paper, we propose a novel closed loop elastic structure which can generate snap-through buckling with bending and twisting. The proposed mechanical closed loop structure consists of two rotary shafts and a belt-shaped elastic material which two ends are connected to one active rotary shaft and one free joint type shaft, respectively. Driving the active rotary shaft continuously, we can obtain impulsive motion by using snap-through buckling repeatedly. The change of the driving torque and the angular momentum generated snap-through buckling are shown by numerical simulations. The proposed mechanism is applied to a compact swimming robot which can turn impulsively at a maximum rate of 135[deg/s] underwater. In addition, it is shown that an offset parameter of the proposed structure is useful for generating impulsive turning motions of the compact swimming robot effectively.
This paper discusses a nonprehensile dynamic manipulation of a deformable object, where the shape of a sheet-like rheological object is dynamically controlled by the combination of the inertial force and the frictional one generated by the plate's rapid motion. We first introduce a linear viscous model for approximating the object deformation characteristics, focusing on the final shape of the object. Assuming that the plate has two degrees of freedom: a translational motion and a rotational motion, we derive two sufficient conditions to deform the object: one to enlarge it and the other to contract it. Then, we show the cyclic plate motion leading to the continuous object's deformation, and finally simulation and experimental results are shown in order to verify the proposed method.
A new robot hand dynamics model is proposed, in which rolling constraints are incorporated and orderless relation between two actuations of a saddle joint is taken into account. For realizing a dexterous robot hand, it is necessary to imitate the orderless actuations because a saddle joint of human being plays a vital role to perform thumb opposability. Arising around the opposition axis connecting the two contact points of each finger-tip with an object, spinning motion is treated in a faithful manner by introducing a physical model of viscous force around the spinning axis of the object. A class of control signals without object kinematics or external sensing is used. It is shown that the two trajectories of the models with ordered joint actuations and the proposed model with the orderless ones are different through numerical simulations. The difference demands to consider the orderless joint model as a saddle joint model. Any solution of the colosed-loop dynamics converges to an equilibrium state establishing force/torque balance through numerical simulations based on the derived model.
This paper proposes a robot which can play air hockey game with a human. The robot consists of a 4-axis robot arm and a high-speed vision, and the robot is controlled based on visual information at a rate of 500[Hz]. The robot system has the abilities to adjust the strength level and to change the strategy based on the game situation. A system designer can easily adjust these abilities by setting several specified parameters. In this paper, first, a recursive trajectory generation using continuous images from high-speed vision is explained. Secondly, the response control to adjust the strength level of the robot is explained. Thirdly, the decision-making using AHP (Analytic Hierarchy Process) is proposed. This ability enable the robot to switch the game plan. Finally, we show the data of experiments and verify the effectiveness of the system.
Pedestrian detection is one of the key technologies for autonomous driving systems and driving assistance systems. To predict the possibility of a future collision, these systems have to accurately recognize pedestrians as far away as possible. Moreover, the function to detect not only people walking but also people who are standing near the road is also required. This paper proposes a method for recognizing pedestrians by using a high-definition LIDAR (light detection and ranging). Two novel features are introduced to improve the classification performance. One is the slice feature, which represents the profile of a human body by widths at the different height levels. The other is the distribution of the reflection intensities of points measured on the target. This feature can contribute to the pedestrian identification because each substance has its own unique reflection characteristics in the near-infrared region of the laser beam. Our approach applies a support vector machine (SVM) to train a classifier from these features. The classifier identifies the clusters of the laser range data that are the pedestrian candidates, generated by pre-processing. A quantitative evaluation in a road environment confirms the effectiveness of the proposed method.