Радиопромышленность | 卷:28 |
ABOUT METRICS OF SELF-SUFFICIENCY AND INTELLIGENCE OF ROBOTIC COMPLEXES AND CYBERPHYSICAL SYSTEMS | |
O. O. Varlamov1  | |
[1] Bauman Moscow State Technical University; | |
关键词: artificial intelligence; robots; mivar; mivar networks; reasoner; autonomy metrics; intelligence metrics; decisionmaking systems, robotic complexes; cyberphysical systems; behavior; computational complexity of algorithms; intelligence; autonomy; active reflection; logically solving systems; expert systems; | |
DOI : 10.21778/2413-9599-2018-1-74-86 | |
来源: DOAJ |
【 摘 要 】
The common feature of cyberphysical systems (CFSs) and robotic systems (RSs) is availability of control systems. CFS is a more general concept, which includes also robots. There are several levels of management of the CFS, starting from the reflex level. For advanced self-contained s RCs the appropriate decision-making systems (DMSs) are designed at the logical level, in which logical inference and automatic construction of algorithms from the «If-then» products, corresponding to the elementary actions of the RTCs are performed. In the field of artificial intelligence the appropriate expert systems were engaged in a logical conclusion, and such systems are currently called «problem-solving and inference systems» or «reasoners». To evaluate their intellectual level the appropriate metric in the form of «variables rules» is used, and such metric is based on estimating the maximum possible number of combinations equal to the factorial of the number of decision rules (products). In this case, the evaluation criterion is «the worst case of complexity» with a complete search of all possible rules describing the situation. The autonomy of the RCs is determined by the intelligence of their control systems, which is expressed in the RC ability to make decisions in various situations. As a metric for evaluation of the intelligence of CFS and RC, it is proposed to use the metric of «variables rules». It is demonstrated that due to the new data representation and the linear computational complexity of the logical inference (without a search!), mivar reasoners make it possible on the ordinary computers to resolve the task of situational control for CFS and self-contained multi-level RCs of multiple purposes and deployment types with a speed of more than one million production rules per microsecond.
【 授权许可】
Unknown