Project goal is enhancement of researches of human cardiovascular system. Mainly we focused our efforts on problems of control of cardiovascular system by means of specific Sections of nervous system. The idea of the project was to join mathematical model of cardiovascular system developed at INRIA with neuron-like model of control system developed at ISP RAS. Some properties of AAC system corresponds to properties of nervous system. Joining of these two models produces autonomous model of cardiovascular system that may be used, for example, for diagnostics of some pathological processes. In the context of this project two-tiered control system that simulate two main levels of control in cardiovascular system was developed.
AdCAS (Adaptive Control of Active Suspension) system has been developed during 2000 – 2001 years as a part of the project initiated by ATS Soft. AdCAS system is based on AAC method, it provides automatic extraction and accumulation of knowledge about changing properties of oscillation motions of a vehicle (for example, of the car). As AdCAS adapts to specifc vegicle it becomes able to set car body position into target position. Target functions are determined by developers and may provide demands on increasing of moving comfort (smoother moving), increasing stability and control quality. As means of changing car’s body position it is possible to use various actuators, for example, shock-absorbers with variable viscosity liquid (MRF) or hydraulic-pneumatic systems. Developed software model of this system showed effectiveness of adaptive control.
System Pilot has been developed our department as part of the project with Lavochkin production bureau and ÖÍČČĚŕř. This project represents software prototype of control system for satellite orientation based on AAC methodology. The system automatically extracts and accumulated knowledge to control the object in the accordance with target functions. Target function may be defined, for example, as keeping the satellite angle position in the given range position with the given accuracy. Experiments performed with software models developed has shown that application of this system causes control quality increase in 5-7 times.
System "Tactician" has been developed in 1995 as part of the project held by ISP RAS for Russian Federation President’s analytical center and with President’s Program Center. This project had examined ability of applying AAC methodology for supporting decision making for social objects control. As the result of the project we’ve showed that with cooperation of specific data representation AAC methodology may be used for developing mentioned adaptive control systems. We found interesting decisions of initial learning of the control system. The project also demonstrates efficiency of this system on software models.
As results of this project we proposed alternative principles of patterns theory that are based on "Autonomous Adaptive Control (AAC)". methodology. We developed new informational structures for presentation empirical knowledge inside the control system about controlled objects with non-linear properties: a) table forms, b) tree-like structures. We developed methods for automatic knowledge base update during single process that combines both learn and control processes. We developed principles of developing control system based on neuron-like approaches that correspond to AAC methodology. We developed few models of formal neurons. We provided research on methods for developing neural networks for such subsystems as pattern formation and recognition system, knowledge base, decision making system and others. We developed object-oriented IDE CASE system “SPINS”. This CASE system has been designed for developing neuron-like realizations of control systems based on AAC method. We provided modeling of neural networks of AAC subsystems using SPINS. First version of "CAPTAIN", – multipurpose adaptive control system was developed. This system is dedicated to control objects that may react dynamically in non-linear oscillating manner.
In the course of this project we investigated ways of combining AAC methodology with some new methods that are used in pattern recognition systems, control systems and search engines. The following principles of integration in the AAC systems were developed:
- fuzzy logic systems,
- determined chaos systems,
- artificial neural networks,
- genetic algorithms.
Project N 128.8 Ministry of science and technologies, "Developing neuron-like realization of autonomous adaptive control system". Direction 05.04128 IČČI, N gov. registration 01.9.70 009436. 1997 – 1998.
AAC methodology have mission for modeling natural nervous systems. So, we find appropriate methods of realization AAC systems. In particular, special attention in the group is given to developing AAC systems based on specially developed models of neurons.
Goal of project is developing and research of prototypes of controlled objects that are based on developed methods and technologies for developing autonomous adaptive control systems.