When a humanoid robot works with a human, it is needed for the robot to pass and receive objects. The robot has to detect that the laden objects are operated by the human on the whole body in object passing tasks. It is important that the robot detects not only push but shear generated by pulling action and twist generated by rotating action for the robot to detect human operations. In this paper, we dispersedly arranged 347 multi-axis soft tactile sensors to a humanoid robot. The multi-axis soft tactile sensors are small muti-axis sensors with urethane structure and they can be placed densely on body of a humanoid. We calculate a deformation vector from the deformation of sensor's upper surface using the output of a muti-axis soft tactile sensor. Next, we suggest the method to detect deformation state in contact surface consisting of soft tactile sensors using deformation moment and average of deformation vectors in the contact surface. Finally, we confirmed the usefulness of the fullbody tactile sensor cover and contact state detector by object passing experiment between a human and a robot.

View full abstract

The 3D reconstruction of the terrain by airbone LRF (Laser Range Finder) requires the 3D motion of sensor, however the payload of small airbone platforms, such as radio control models, are highly restricted, and precise inertial motion units are too heavy to be mounted on it. The HDL32E, an LRF recently provided by Velodyne, is light enough to be installed to the airbone vehicle such as a blimp, and has 32 rays for scanning so that it can sense wide range of terrain at one scan cycle which means low dense 3D pointcloud are collected at very short time. This paper proposes the fast and simple method for integrating those low dense pointclouds to high dense pointcloud without any intertial motion sensor. The method uses equal or more than 3 planes not parallel on the terrain, and estimates LRF's 3D motion. The final integration result will be shown after the optimization of laser scanner's motion.

View full abstract

In this paper, a fast and stable motion controller for a robot with a flexible arm using IDCS control scheme is proposed. The IDCS controller scheme, corresponds to a numerical computation of the inverse dynamics of the systems, allows stable high performance control even when the system model is inaccuracy. We can get good simulation model of the robot for the IDCS by estimating inertial parameters using generalized coordinates of the baselink, the joint angles and the external forces information. We applied IDCS with the controller adequate for suppressing vibration: Dual Model Matching (DMM) to the lift table and consider the reference tracking of IDCS with reduction of the vibration of the flexible arm.

View full abstract

In this paper, an analysis of contact forces in quasi-static manipulation under rigid-body mechanics is studied. To evaluate the contact forces, we introduce a modified version of constraint on friction forces, which was originally derived for statics of power grasps by Omata et al. Furthermore, using the constraint, we also present a robustness measure for quasi-static graspless manipulation such as pushing and tumbling. The measure evaluates how much the object can resist external disturbances without changing its motion. We can calculate the measure approximately by solving linear programming problems. We show some numerical examples to prove that our proposed method can evaluate robustness of graspless manipulation more accurately than previous methods, and experimentally verify the calculation results.

View full abstract

This paper proposes a method to construct the accurate and dense three-dimentional (3D) model of an outdoor scene by a rotational stereo camera. The rotational stereo camera system consists of two binocular stereo camera which are a horizontal stereo camera and a vertical stereo camera. The vertical stereo camera can change angle whose range is from −180[deg] to 180[deg] rotating the axis back and forth and can capture the target at various angles. It is difficult to reconstruct outdoor scenes accurately using one binocular stereo camera because the distance to a target is very far compared with the stereo camera's baseline length. In this paper, binocular stereo cameras are treat as a multi view stereo. But a corresponding method is required to match between very different images. Therefore the proposed method uses Epipolar transfer as a corresponding method. Epipolar transfer can match images without using Template matching like NCC. As an experiment, a blimp robot with a rotational stereo camera is used to capture aerial stereo images of the ground. The proposed method achieves the accurate reconstruction of the outdoor 3D model.

View full abstract

Recently, flagella and cilia have gathered attension since their important role in mammal cells and relation to genetic diseases of human were revealed. Chlamydomonas is one of the model organism of flagella and phototaxis research. The purpose of our research is to track a freely swimming individual Chlamydomonas cell to observe its flagella and whole body movement with dark field microscopy to investigate the phototactic reaction of Chlamydomonas. In this paper, a method for three-dimensional tracking of Chlamydomonas using high-speed visual feedback in dark field microscopy is proposed. The method was implemented to an existing microscope tracking system. It demonstrated successful three-dimensional tracking of a swimimng Chlamydomonas for several tens of seconds. The precisions of the 3D position measurement were estimated as ±1[μm] (XY) and ±10[μm] (Z).

View full abstract

This paper proposes a method of lane marker detection for platooning. In order to reduce air resistance, it is desirable to shorten the distance between two vehicles. If the vehicular gap is very short, conventional methods, which detect lane markers ahead in images captured from a front camera, are useless because lane markers are occluded by a vehicle in front. To solve this problem, the proposed method recognizes a lane markers in images captured from two downward side cameras equipped with the front side and the rear side of a vehicle. First, candidate points of lane markers are extracted in each image by edge pair detection. Then, straight lines representing lane marker are detected by applying Hough transform to these candidate points. A lateral position in a traffic lane is estimated from a position of a straight line in an image of each downward side camera. Because a downward side camera can take only a narrow area directly under it, lane markers must be detected from short parts of them. Therefore, the proposed method uses a multi-information map containing lane marker information to detect lane markers. The proposed method has been implemented in the image processing hardware whose CPU satisfies on-vehicle specifications. Experimental results show the effectiveness of the proposed method and a lane detection device. In experiments conducted at the test course, four platooning vehicles, which consist of three heavy-duty trucks and one light truck, ran automatically at 80 km/h, maintaining 4m of distance between vehicles.

View full abstract