This project improves "Autonomous Adaptive Control" (AAC) method. AAC method provides approach for building control systems that exhibit both recognizing and control features in a single process that is based on biological models. To increase effectiveness of AAC systems and to broaden areas of practical applications it is necessary to include multilevel hierarchical control ability into AAC methodology. Such requirement and corresponding possibility may be inferred from principles of organization and functioning of AAC control system. That possibility comes from principles of organization and behavior of AAC systems and concerned with ability of automatic aggregation of information elements, that are operated by control system "patterns" and "knowledges". We expect that this approach will give ability to develop multi-level hierarchic control without considerable increasing structure of control system.2. Project RFBR # 04-01-08023-îfi-a (2004-2006 years) "Developing multilevel neuron-like system of autonomous adaptive control for mobile robot".
Control system for mobile robot has to exhibit adaptability of control on the lower (actuator) and on the middle (tactic) levels. We expect to solve the following problems of automatic adaptation:
- automatic elaboration of stereotypes of behavior for avoiding the obstacles,
- memorization and automatic correction the map of the environment,
- associative memorization of regions and details of environment,
- automatic recognition of targets in environment,
- adaptive routes discovery to destination point.
Mobile robot developed for the project shall have both research value (ability to research and present various algorithms of control) and applied value (similar to various applied problems). Control system of the robot is designed to be neuron-like and could be possibly implemented by neural computer.
Designer 4GN - is toolbox for prototyping control systems based on Autonomous Adaptive Control (AAC) method. By means of 4GN it is possible to develop models of control systems for various domain areas in short periods with minimal costs.
Primary features of Designer 4GN:
- independency from problem domain,
- explicit decomposition of control system into primary functional blocks,
- decreased control system development time and costs,
- logic design and development for base components,
- ability for integration with existing tools (MATLAB) and with models that were developed by means of these tools,
- flexible visual-oriented interface to develop control systems,
- built-in editors for neural network components,
- flexibility: ability to modify functionality of the system on any level,
- extensibility: ability to add new functionality to existing projects,
- ability of integrating with databases via OLE DB,
- ability to develop own component for visualization.
Application domains Designer 4GN:
- Car suspension systems
- Railway suspension systems
- Navigation of mobile robots
- Adaptive control of prosthetic devices
- Orientation of artificial satellites
- Control of network capacity
Developing of adaptive control systems based on methods and approaches that were developed at Autonomous Adaptive Control Methods Department ISP RAS for robotic devices. Considered methods and approaches include genetic algorithms, fuzzy logic and elements of determined chaos.5. Project "Adaptive control system for bioelectrical prosthetic device of upper extremities on the base of bionic Autonomous Adaptive Control method in the context Program of fundamental researches Presidium RAS "Fundamental sciences for medicine"". 2005 year. Joint participant Albrecht Insitute of prosthesis (St. Petersburg).
This project deals with developing of adaptive control system for prosthetic devices of upper extremities based on bionic "Autonomous Adaptive Control method". Control systems designed according to the method are capable of learning and re-learning at the same time. Such control systems are able to deal with coarse mechanical models of controlled objects. Such decent requirement is very important for complex prosthetic devices, because many parameters of the system "human-prosthetic device-environment" are changing constantly and it is nearly impossible to formalize them. Under this condition application of AAC-based control system may become a crucial point.6. Enterprising project "Adaptive control system for program model of mobile robot "Gnom # 8"". Project has begun at 2001.
The aim of the project is development of mobile robot based on Autonomous adaptive control methodology. Control system of the robot has to be implemented as an artificial neural network that must exhibit properties of biological nervous system.7. Enterprising project "Modeling and researching of behavior act in living organisms and program biological models". Joint participant K.V. Anokhin group, Insitute for normal physiology RAMS.
Researching the nature of behavior act by comparison of experimental results on living beings (mice) and software models of mobile robots that were put in the same environment.