OMRON's core technology, Sensing and control, is intended to extract high-value information from any type of phenomena and output it in an intelligent manner for subsequent human or system processing. OMRON's Knowledge Information Control Technology is an essential component that makes this possible.

In the field of Knowledge Information Control Technology, OMRON's ongoing research is concentrated in three areas: extraction of characteristic elements, learning, and inference. These three capabilities make it possible to extract valuable information from sensor-detected data such as images or sound, and make inferences from the results of the extraction. A very advanced level of knowledge and expertise is required to draw a certain conclusion from specific information, but this is exactly where machine learning comes in by putting the knowledge and expertise into the form of machine-executable data.

Although the concept of Knowledge Information Control Technology at first seems difficult to understand, it is in fact already incorporated in many of our daily-use products. Take the example of an anti-theft system for automobiles. Such a system uses a sensor to detect sound or vibration generated around the car, determines a possible theft from the detected data, and triggers an alarm. The system can infer that the car is in danger of being robbed from a certain characteristic factor-sound. Learning is the process that lets the system accumulate knowledge so that it can correctly identify a certain sound as a dangerous situation.

Learning ability is essential for automating machinery. The capability to determine which of several extracted characteristics is most suitable for making a judgment can shorten the time required to make an inference. Learning how each type of data should be interpreted is vital for drawing correct conclusions.

There are two methods by which machines can learn. The first method is learning from data. By accumulating data that shows which extracted information should be accepted or rejected, then learning from it, machines will be able to increase the accuracy of the inferences they make. Fortunately, this is the area in which machines are most capable-analyzing the distribution of voluminous data in a short period of time.

* Learning from data

The other learning method is to incorporate human knowledge and experience. When a machine encounters a situation it has never experienced before, the machine is unable to determine which data it should refer to. In these cases, machines can be "educated" by inputting the results of expert judgments based on human knowledge and experience, and building a knowledge database based on these judgments. In this way, machines can be taught how to perceive a certain situation as humans do based on visual or auditory sensations.

*Learning from humans Inputting the results of human judgment in the form of data that machines can manipulate.

Through repetition of this training process, machines can learn to instantly assess even complex situations involving many intricately entangled factors-an ability that was once impossible to achieve without advanced human skill and judgment. Even in cases where a machine is unable to make a judgment, it can notify a human operator, thus making it easier for the operator to determine what the machine needs to be taught.

OMRON has developed a wide range of sensing technologies, and during the course of development, it has accumulated and analyzed an immense amount of data. By combining this accumulated data with knowledge data based on the experience of experts, OMRON's Knowledge Information Control Technology aims to allow machines to make more appropriate decisions using less information.

Ensuring Production Quality

Return to Knowledge Information Control Technology Top