A NEW METHOD OF EVOLUTIONARY MODELING USED IN REAL-TIME SIMULATION

Genetic programming （GP） has a strong ability in model-building of complex systems, and can find a series of correlative function or differential equation to describe the static or dynamic process of simulation systems, so it is an effective means of describing complex systems. But GP has the following disadvantages: the efficiency is low and the precision of models isn’t high, so a new evolutionary modeling algorithm （two stage evolutionary modeling, TSEM） has to be proposed. In the TSEM algorithm the system bulletin of the AI is introduced to proclaim some better solutions, so the genetic operating utilizes the bulletin information. The TSEM algorithm is used to a real-time system: marine engine system, and utilize the characteristic of a simulation system to guide the GP search and reduce the searching blindness. The experimental results show that the TSEM algorithm is better than the traditional GPs in speed and precision of models.