Research On The Software Reliability Of Car Navigation System

Nowadays the development of navigation system for land vehicles is just in theascendant. It has been widely used in many fields such as physical distribution andintelligent transportation system. With the development of the navigation system technique its scale is larger and larger and its function is more and more important. People put forward a higher request against the software of the integrated navigation system. Not only it is requested that the software of the integrated navigation system renew gradually and adopt the design method of optimizing it, but also people pay more special attention to the high software reliability of the integrated navigation system on the basis of fulfilling its basic possessing function. This paper is based on the theory of software reliability engineering and regards ensuring the high software reliability as the aim of research against these new requests of the integrated navigation system and carry out the forecast of the software reliability.In allusion to the integrated navigation system software's characteristic the paper uses the establishment software reliable model the method to conduct the forecast research to the software reliability. Through research and analysis of traditional software reliability models and neural network structure and characteristics we pointed out their deficiencies. The low speed of learning is a main problem for the BP neural networks in fault diagnosis systems. On the basis of others theory and results of software reliability prediction on neural network, the momentum term and variable studying rate are increased to the standard BP algorithm, and designing optimally of the network structure and numbers of the hidden layer nodes.After then we use the improvement BP neural networks carries on to the integrated navigation system software disaster number static state, the performance forecast and the trundle forecast. And further in the comparison confirmation classical reliable model and in the BP neural networks model respective characteristic and the insufficient foundation, the synthesis classical picture and the BP neural network's superiority forms one kind of mixed model, carries on the training using the BP algorithm, and confirms it after the actual test to have compared to the sole classical picture and the BP network higher forecast precision.