Simulating reliability of the robotic system software on the basis of artificial intelligence

The evaluation model of the software reliability is implemented on the basis of artificial intelligence methods, using artificial neural networks. On entry of the model the debugging time is served, on return the outlook the value of the failure rate is formed. For the model implementation a special type of neural network — a vertically-layered neural network is worked out. The model accuracy is increasing leads up to the buildup of layers in the neural network. The implementation of an artificial neural network of this type of multi-layer modular way is considered. The main results are protected by the RF patents for inventions and useful models. This article is devoted to research of the complex software reliability of control systems of different functions and is based on the previous researches of the author in this area. In the article the following questions are considered: 1. Adaptation of earlier published model for an assessment of reliability of the procedure oriented software to the object-oriented software. 2. Reduction of a known formula of predicting model of reliability to a look convenient for creation of an artificial neural network. 3. Creation of an artificial neural network for forecasting and an assessment of complex software reliability. 4. The description of a new way of realization of the constructed neural network. The main results of the work described in article are protected by patents for inventions and useful models in Russia.