This paper shows analysis results of dialogue histories between chat robot and eldery peoples. The chat robot have been developing for the purpose of long-term use as home robot. The authors tested the chat robot with eldery peoples who are over 65 years old in a week to 10 days. During the experiments, the participants chated with this robot anytime they want in their daily life. The authors analyzed this dialogue histories. Therefore, the dialogue includes around 30% noun. The recognition rate is upward tendency in each day, but the rate is depends on the each user's speech condition. Moreover, the participants talk well about “go-out” and “meal”.
Jumping Scouter is a rocket-powered robot for planetary exploration. The light-weighted robot traverses long distance, independent of terrain conditions. For example, it theoretically jumps 43[km] with only 290[g] of vehicle mass and 98[g] of black powder. It also acquires landscape imagery of wide areas during and after traversal jumps. This paper describes the robot's concept, system configuration and primary evaluation tests for jumping functionary.
This paper addresses a framework of motion transition for humanoid robots by switching various types of controllers. In motion transition, one of the key issues is the physical constraint in humanoid dynamics. In this study, by switching controllers based on the maximal output admissible (MOA) set, we realize motion transitions without exceeding the physical constraint. In particular, we present a novel computation method of the MOA set for a limit cycle controller, which is often used for steady walking. By approximately calculating the MOA set via sample point cloud, we can obtain a closed form of the MOA set even in a nonlinear system. We demonstrate examples of motion transition and verify the validity of the proposed framework.
A purpose of this study is to achieve wrapping operation by robots. As a teaching method of wrapping operation, this study uses an intuitive instruction based on a movement of a hand, and generates robot commands from the instruction. This study represents wrapping operation based on a three-layered model as intermediate representation, which is originally proposed in our previous study. In this paper, as a key elements of this study, we describe the teaching method which is a part of inputting to the model. This method can understand an intention of the rough instruction and generate appropriate intermediate representation which represents the intention. In the intermediate representation, a part of hand path generation method is extended to loosen limitations about shapes of objects and fabric that can be handled. Finally, we integrate them into a total wrapping robot system, and confirm that the proposed method is effective to achieve wrapping operation by a robot.