As has been claimed by Barlow , and reported by some recent neuro-physiological researches, at higher levels in the hierarchy of representations in the brain, sparse coding is adopted. Sparse coding is a kind of neural representation in which a very small number of neurons fire selectively. Because of the small overlaps between the firing patterns, the codes have the property of uniform metric, which may correspond to abstract symbolic representation of physical patterns . We propose here a system that generates sparse codes of concepts of motions from accumulated feature vectors of observed motion patterns, by extending our previous research . We propose an associative memory dynamics model with a self-organizing nonmonotonic activation function, which automatically finds out the hierarchical cluster structures in the stored motion patterns. Based on analysis of the dynamics of this model, we design an output function for the attractors, which can generate the sparse codes of the symbols of motion patterns.
This paper presents a method of selecting landmarks for real time robot navigation. We first build a 3-D landmark map of an unknown environment with trinocular vision, and then the robot with monocular vision localizes the own position by matching SIFT keypoints extracted from the input image to the landmarks based on their visual appearances. It requires much computation to detect visual features corresponding to all the visual landmarks. There are many redundant landmarks among them and they can be excluded in the calculation without major degradation of localization. We propose to select useful landmarks considering their contributions to the localization accuracy derived from the analysis of geometric error propagation, and both the reliability of detection and matching of the corresponding visual features. We compare the proposed method with the one which uses all the landmarks by experiments in real indoor environments. The result proved that the localization accuracy of the proposed method is almost the same, and the computation time is greatly reduced.
This paper discusses the kinematical analysis of parallel mechanisms using the multi-wire driven method. This method allows for the use of two or more wires attached to one actuator. One type of this method can produce passive constraints, whereas another type can produce active constraints in an area where a wire does not exist. The definition and the features of this method are then described. Then an analytical technique to obtain the number of degrees of freedom and possible directions of motion is proposed. Its validity is verified using an example.
In this research, a small crawler-type searching robot driven with the pair of flexible shafts is proposed. It is capable of being effectively utilized for a search inside of rubble to quickly search for victims when a disaster such as an earthquake occurs. Since the crawler robot does not have to mount any power unit on the body, it is possible to construct the body of relatively light weight. Furthermore, since any electrical power source is not mounted on the body, the crawler robot can perform a searching operation under a sodden environment without additional waterproofing. To construct the proposed movement mechanism, we have to devise the structure of the flexible shafts. The flexible shafts are formed by inserting torque transmission driving wires (inner shafts) into tubes (outer tubes). In this paper, we investigate the characteristics of the outer tubes and the torque transfer characteristics of the flexible shafts. Then we improve the robustness of the outer tubes and the structure of the flexible shafts to enable a high-speed movement of the crawler robot.
This paper proposes an object movement detection system covering large areas of a room by using multiple cameras. When object movement detection for whole of a room is performed, there are several challenging difficulties: sizes of objects on the camera images are small, non-objects such as humans also exist on the images, objects are sometimes difficult to detect in the specific viewpoints because of occlusion by humans or furniture or color similarity to near objects. In this work, we propose an object movement detection method by integrating multiple viewpoints via features extracted from “stable changes” on each viewpoint. To discriminate whether object or non-object, we focus on motion of changed regions. Experiment in a room environment shows the multiple view integration method with the color and position features improves recall rate of object detection performance.
In this paper, we propose a running gait generation method for a biped robot using the dynamics model with two foot masses, an inverted pendulum and a flywheel. The inverted pendulum and flywheel represent the horizontal motion and the bending motion of the upper body respectively. Initially the acceleration of the inverted pendulum is determined to satisfy the desired ZMP. If the horizontal ground reaction force exceeds its limit, the acceleration of the inverted pendulum is modified and the flywheel is accelerated to compensate this modification. To guarantee the continuity of gait generation, we extend previously proposed concept ``the divergent component of the motion'' to be applicable for the inverted pendulum under time-varying ground reaction force. Using this concept as the boundary condition, the current gait converges to the next cyclic gait. Compared with the boundary condition using position and velocity, proposed boundary condition requires smaller amount of modification of the desired ZMP trajectory.
Recently the necessity of autonomous or highly functional agricultural equipments is increasing because of decrease and aging of labors for farming. The development of autonomous vehicles in an orchard is required as one of the equipments because an orchard is usually on a hill or mountain where it is very tough for farmers to work. In this research, a new design method of an unmanned ground vehicle (UGV) in an orchard is proposed by using image-based control with a central catadioptric camera. A central catadioptric camera is very effective to keep target objects in the camera field of view because of its wide area view. The effectiveness of our proposed method is confirmed by experimental results in an orchard.