Yasuto Tamura

A Decision Making for a Robot Based on Simple Interaction with Human

Recently, an intelligent robot is expected to operate our living area. To realize, a robot should make a decision for action from a simple order by human. For decision making of a robot, it is important that a perception of environmental situation and adaptation to the preference of a person. We have proposed the learning method based on SOM to adapt environmental situation and the preference of human. Through experiments by simulation, we verified that the proposed method can consider the changing of attribution by time variation. And, the decision making of a robot can be adapted to the preference of a person through interaction with a person.

Path Planning Based on Direct Perception for Unknown Object Grasping

This paper discusses a robot path planning based on the sensation of grasping for a service robot. The previous method should recognize the accurate physical parameters to grasp an unknown object. Hence, we propose the path planning by the sensation of grasping for a decreasing of computational costs. The sensation of grasping affords the possibility of action to a robot directly without inference from physical information. The sensation of grasping is explained by an inertia tensor of three-dimensional point cloud and a fuzzy inference. The proposed method involves an unknown object detection by a depth sensor, and the path planning based on the sensation of object grasping that is determined by the characteristic of the robot and the state of the object. As experimental results, we show that the robot can grasp the unknown object which the robot arm cannot reach on the table due to the movement to the position suitable for the object grasping.

Development of pipe inspection robot using ring-type laser

This paper describes the robot capable of inspecting a pipe. The developed robot is composed of three bodies, six legs, a laser range finder, a ring-type laser, and a CCD camera. The six legs are composed of three worm gears that are arranged at an angle of 120 degrees to run inside pipes of different diameters. The robot is designed with two joints to change the direction in pipe junctions, and the robot can measure a fore pipe shape using the laser range finder. The ring-type laser and the CCD camera are installed in the front the robot to detect and evaluate defects of a pipe. The CCD camera takes an image of the ring-type laser light in a pipe, and defects in a pipe are detected and evaluated from a break of the ring-type light by the image processing. We verified the effectiveness of the robot through the movement experiment and the defect inspection experiment.

Gait pattern generation under disturbance force

This paper describes a gait pattern generation and the mechanisms of a biped humanoid robot that has 42 DOFs in total. The movable angles of the joints of the biped robot are about the same as those of a human. Its height is 1.66 [m] and its weight is 62.5 [kg]. If disturbance forces are applied to the waist when the robot is walking, the robot should change its step length to control its balance and not fall down. The gait pattern consists of a leg trajectory, an arm trajectory, and a waist trajectory is created using a fuzzy control rule. A waist trajectory that can cancel the moments generated by the motion of the legs is also determined by using the preview control. To confirm the efficacy of the gait pattern, walking simulations were conducted using MATLAB/Simlink and ADAMS.

Object recognition using multiple instance learning with unclear object teaching

We propose an object recognition method for service robots under the constraint of uncertain object teaching by humans. In previous object recognition methods, the training phase required a large number of prepared images and also required the training data to not have a complex background. However, for robots to perform daily tasks, they should be able to recognize objects despite unclear object teaching by humans. In order to mitigate the effect of features in the background on object recognition, our proposed method classifies local features based on saliency from video images. In this paper, we demonstrate the efficacy of the proposed method in recognizing target objects despite unclear teaching by the user.

Object detection based on saliency map using reference image containing complex background

We propose an object detection method based on a saliency map using a reference image containing complex background for service robots. In previous detection methods, images that the user prepares in advance contain mostly simple background. However, in order to make robots perform daily tasks, the robots should be able to detect an object using snapshots that contain a complex background. In order to decrease the effect of features in the background, our proposed method classifies local features based on saliency from images. This paper shows the efficacy of the proposed method; furthermore, we demonstrate that our service robot detects certain objects according to the proposed method.

Object recognition based on inhibition-of-return control using particle filter

We propose an object recognition method based on a saliency using a reference video containing complex background for service robots. In previous object recognition methods, in training phase, images that the user prepares contain mostly simple background. However, in order to make robots perform daily tasks, the robots should be able to recognition an object using teaching scene that contain complex backgrounds. In order to decrease the effect of features in the background, our proposed method classifies local features based on saliency from video images. This paper shows the efficacy of the proposed method; furthermore, we demonstrate that our proposed method recognizes certain objects.

A direct perception for decision making of a service robot

Recently, an intelligent robot is expected to operate our living area. To realize, a robot should make a decision for action from a simple order by human. For decision making of a robot, it is important that a perception of environmental situation and adaptation to the preference of a person. We have proposed the learning method based on bottom-up process using SOM and top-down process using Fuzzy Inference. Through experiments by simulation, we verified that the proposed method can consider the changing of attribution by time variation. And, we discuss about features of the different proposed method through comparing.