Hierarchical growing neural gas for information structured space

Proceedings of the Institution of Mechanical Engineers, Part I:

2011

This paper discusses human-friendly tele-operation for the next generation using robot partners through sensor networks. The emerging synthesis of robot technology, information technology, network technology, and intelligence technology based on cognitive psychology and brain science can be a key approach to realize human-friendly tele-operation. In particular, the strong link between collection, structuralization, and visualization of information through these technologies interacting with human interest and intention is the most important. First, this paper looks at the current research studies on tele-operation. Next, several examples are shown of human-friendly tele-operation of robot partners. Finally, the various types of technologies required for human-friendly tele-operation for the next generation are discussed.

Section: Sensor Fusion For Information Extraction (4)mentioning

confidence: 99%

Human-friendly tele-operation for robot partners

Proceedings of the Institution of Mechanical Engineers, Part I:

2011

“…In this result, LGAC detects a person, a red ball, and blue bucket efficiently. Fig.7 (b) ~ (d) shows a preliminary experimental result [14] of HGNG where the number of layers is 3 (l=3). The lowest layer of GNG efficiently covers the sampling data, and the second layer of GNG covers the spatial distribution of nodes in the first GNG in the middle level.…”

Section: A People Tracking Based On Lgacmentioning

confidence: 99%

mentioning

confidence: 99%

Topological environment reconstruction in informationally structured space for pocket robot partners

2009 IEEE International Symposium on Computational Intelligence in Robotics and Automation - (CIRA)

Yorita

2009

Self Cite

Thispaper deals with environment reconstruction in informationally structured space used for pocket robot partners. The environmental information is measured through the sensor network, and is stored in the remote host computer. The pocket robot partner accesses the environment information through wireless LAN, and receives a 3D model of the environment including people and other robots. Therefore, we propose a topological environment reconstruction method to extract the state of the environment. Finally, we show several experimental results of the proposed method, and discuss the effectiveness of the proposed method.

“…An artificial neural network is one of promising methods to realize machine learning with or without feedback error functions. Neural networks have been applied to nonlinear control [9]. The neural network realizes a nonlinear discriminated function.…”

Section: B Learning Algorithmmentioning

confidence: 99%

Trajectories tracing for a pitching robot based on human recognition

Yuuki

Yamada

2009 IEEE International Symposium on Computational Intelligence in Robotics and Automation - (CIRA)

2009

Self Cite

In this study, we discussed human recognition method for the trajectory tracking. Visual perception is very important to realize the feature extraction from the time series of images, but it is very difficult to perform the object tracking. We classified errors occurring in the throwing; errors caused by internal factors and errors caused by external factors. At first, we propose a theoretical method of calculating the trajectory of a ball from the set-values. Second, we propose the method of calculating the trajectory from parts positions. These positions are recognized from the image captured by CCD camera. In this calculation, the pattern recognition were used to find parts positions. These positions are the released position of a ball and the position of the fulcrum. By this way, e.g., we can find the released position error of a ball and the speed error of a ball caused by internal factors. We used parabola equations in these calculations of trajectories. Third, we propose the method of extracting the trajectory from the time series of images captured by CCD camera. The specifications of camera are "480 by 360 pixels", "RGB color" and "29 frames per second". We propose the method of recognizing a position of the flying ball from images of the movie, directly. The robot plots the trajectory of a flying ball. By using this method, we can find the errors caused by external factors, e.g., we can suppose the influence of the air resistance working to the ball. Finally, we performed the experiment of the trajectories tracing for a pitching robot based on human recognition.