₪ … who are we?

   IngB RT&S is a specialist company for bionic processes, which means nothing other than that it decodes biological methods of signal recording (sensor technology), meaningful signal processing (feature extraction) and knowledge representation as the basis for the recognition of sensor data representations (situation identification, situation categorization) and converts them into an algorithmic representation (machine intelligence).

   This simple-sounding approach, however, has a major hurdle: since nature can do many things but not count, IngB RT&S  had to develop completely new methods and processes that had to be potential-oriented on the one hand and symmetry-related in their dynamics on the other.

   In the course of its existence, IngB RT&S succeeded in doing this in its two current main areas:

• machine vision

and

• self-adaptive decision support systems.

   In both areas, IngB RT&S 's applications have been recognized with numerous national and international innovation awards.

  Let us briefly describe the unique selling points of both areas:

I ) Machine vision:

   Classic or AI-based optical filtering methods do not guarantee constant contour recognition or reproduction of true color impressions in camera operation, since, unlike biological image processing by the eye, they do not self-adapt to lighting situations/changes in lighting. As a result, the processed image sequences appear noisy, the object-defining contours are not constant. This makes automated object recognition difficult or impossible.

   On the other hand, if you work with potentials and symmetries, you get lighting-independent contour recognition that is intrinsically adaptive, i.e. optical filter structures are created that are neither learned nor modified under all possible lighting conditions. They act as robustly and "self-adjusting" as our eyes.

II) adaptive and continuously coding decision support systems:

   If this potential-oriented procedure is transferred to the structure of decision support systems, one arrives at a new type of categorizer that resembles classic neural networks with two layers, but opens up completely new dimensions through the coding method of "computing with activities".

Here too, the following applied:

   When nature has opened up a new, comprehensive processing area, it applies it "at every turn". These categorizers outshine conventional neural-based applications in several ways.

These are:

• New rules are integrated into the old set of rules without delay, without the old set of rules being destroyed or masked.
• Layers not previously taken into account are automatically displayed as "I don't know that yet" instead of being mapped to irrelevant layers
• Their structure is mathematically so compact that the trajectory of the system states can be predicted.
• The new neural systems can also be initially conditioned with little data (states) and grow gradually during operation.

In use today, we look back on a difficult path of R&D efforts that is far from over, as every discovery, every new process, immediately gives rise to at least three new questions.

What was the hardest part of our path? Well, the temptation to integrate classical methods and procedures into our systems and thus leave the bionic path...