Convolutional neural network (CNN) has been successfully used in many fields including image recognition. CNN is composed of input, convolution, pooling, hidden and output layers, and the weights and biases between layers except the ones between convolution and pooling layers are acquired by learning. In comparison to the conventional neural networks, the learning cost of CNN is higher, and the learning time is longer especially when hidden layer(s) are added. Recently, complex- and quaternion-valued neural networks have drawn much attention. In complex-valued neural networks, inputs, weights, biases and outputs are complex numbers, and in quaternion-valued neural networks, these parameters are quaternions. It has been shown that both methods exhibit excellent accuracy in various applications such as classification and function approximation problems with less computational burden. In this study, we propose CNNs with complex- and quaternion-valued neurons where complex and quaternion numbers are used between pooling, hidden and output layers. We here show that CNNs with complex- and quaternion-valued neurons have higher learning ability in handwritten digit image classification with the MNIST data than the real-valued counterpart.

We have proposed a concept of a robotic information home appliance corresponding to one category of home robots, and developed ApriAlpha, a concept model of the robotic information home appliance. ApriAlpha is a wheel locomotion type human friendly home robot which controls advanced home appliances, standing between their users and them as a voice controlled information terminal, and offers security and information services to users. We have integrated various robot technologies such as voice communication, image recognition, planning and motion control on ApriAlpha, and we have introduced a framework of the distributed object technology based open robot controller architecture (ORCA), which we are currently developing in view of its easy extension and efficient development. This paper describes the proposed robotic information home appliance and its concept model, ApriAlpha. The functions of the developed robot are confirmed by performing several demonstrations, and the merit of applying the framework of ORCA to the home robot controller is also confirmed through its development.

We developed a robot system which can play one-on-one volleyball games with a human player. It can not only hit a flying ball back to the player correctly but also have various autonomous functionalities to interact with humans necessary for playing games, i.e., it can pick up a ball of a requested color by recognizing voice instructions, can locate the players by recognizing their faces, and can shake hands. This robot performed the entire game sequence over 100 times at a company-sponsored exhibition, where the ball rally continued over 25 times with an arbitrary player. In this paper, we mainly describe the real-time visual feedback system of the robot which is essential for playing ball games.

We have developed a new unilateral master-slave hand system. The master system consists of a master arm which has a 3-DOF parallel link mechanism and a 3-DOF wrist mechanism, and a dataglove that is usually used for a virtual reality system. The slave system has a 6-DOF arm and a 16-DOF four-fingered hand. In general, a slave system is controlled by position with reference to a master system position. In this system, when an operator makes the slave system work on a complicated task, he must look at the slave carefully and let his arm and hand manipulate discretely. In the new master-slave system, each finger of the slave hand has a fingertip tactile sensor, and it is controlled by position and force. When an operator has knowledge of a task and handling objects and sets the slave fingertip stiffness, the operation can be performed easily by handling the master arm and hand to a rough position.

The authors are developing a multifingered robot hand which is modeled on a human hand. The aim is to identify the technologies necessary for this type of robot hand. The hand consists of 4 fingers, each with 4 joints actuated by DC motors via wires. The hand is attached to the tip of a 6-degree-of-freedom arm. The fingers and the arm are controlled by 6 32-bit microprocessor boards. Each finger has a tactile sensor on the fingertip. This consisting of a silicone rubber cap with a cavity filled with an incompressible fluid. Using these sensors and potentiometers mounted at one side of each joint, contact forces and positions of the fingertips are controlled. Three or four fingers grasp an object cooperatively when each finger is force-controlled using the tactile sensors independently. Evaluation tests of the position and the force control were executed, and experiments requiring the robot hand to pick up a pen and write characters were conducted.