Multiple robot manipulators are used in a coordinated way for carrying out tasks which cannot be performed by a single robot. When two or more robots handle a large object in free space, it is necessary to control the internal force applied on the object as well as the positions of the robot end effectors. In this paper, a learning control algorithm is proposed for the iterative-type position and force control of multiple manipulators carrying an object in free space. The learning control algorithm is of the D-type, and uses acceleration errors and internal force errors for learning. A proof of the convergence of the position and internal force trajectories to the desired trajectories is given. Simulation and experimental results are presented for two-robot manipulator systems to illustrate the effectiveness of the proposed learning controller.
In this paper, a method of identifying individuals using the range image in the 3/4 views is proposed. We have already proposed the method of identifying human faces using the three dimensional facial data. But the method we proposed previously assumed that the three dimensional data for identifying the faces were obtained from only the small region on the facial surface because the three dimensional sensor was set in front of the face. In the proposed method, the sensor is set at an angle to the face so that the three dimensional data can be obtained form the large region on one side of the facial surface. In this paper, we describe the algorithm of identifying human faces using the angled facial data and show some experimental results to verify effectiveness of the proposed method.
This paper describes a method of object recognition based on aspect graph using the Dempster-Shafer theory. The method deals with several sequential images as input images, and extracts basic probability of the model objects for each input image. The basic probability represents ambiguous information for object discrimination. Our approach combines basic probability which is extracted from each image based on Dempster-Shafer's rule for combining. We have applied our approach to the discrimination of objects. The experimental results show higher reliability than conventional Baysian approach.
In this paper, we propose a hybrid neural network (HNN) for identification of unknown continuous time nonlinear dynamic systems. HNN can identify dynamics of unknown systems by using only inputs and outputs. It is the salient feature of our proposal that HNN does not need states of the system nor differentials of outputs. The existence of HNN which approximates input-output relationship of unknown dynamic system within any precision is shown. In addition to HNN, two multilayered neural networks (MNNs) are used as a state observer for HNN and a state feedback controller. Integrating HNN and these two MNNs, a controller for the unknown nonlinear plant is constructed. Simulation results show the validity of our proposal.
This paper proposes a lower visual information processing model detecting a subjective contour which is a kind of optical illusions. In the earlier contributions, a DOG function partial to a specified orientation was used to detect a subjective contour. However, we adapt a DOG function having concentric circles. We consider that a visual information is represented by a position and an orientation in area VI and area V2, since neurons in the hypercolumn have selective responses to a position and an orientation. Our model can detect a positon and an orientation of a line by transmitting one output of a neuron in retina to plural neurons in area Vi through lateral geniculate nucles. We further consider the effect of a fixation point in detecting a subjective contour. We illustrate our proposed model to optical illusions in simulation. The results say that the model can realize neurons which have selective responses for a position and an orientation, and detect a subjective contour depending on a fixation point.