Computer Generated Force is an important research area in military simulation domain. It can be used in combat simulation, new weapon system development planning, and combat efficiency evaluation area. One versus one fighter dogfight combat is very important for air combat training, so, it is also important to modeling human behavior of the one versus one fighter dogfight combat.In this dissertation, we study the human behavior representation technology of short-range one versus one fighter dogfight combat. From the generic framework of human behavior modeling, we divide the human behavior model into two parts: situation awareness and behavior production. The situation awareness module simulates the pilot's situation awareness in the battlefield and the behavior production module generates the combat behavior. We divide the behavior production module into three layers: control layer, tactic planning layer and decision layer. The control layer simulates the pilot's basic maneuvering function. The tactic planning layer plans a series of control instructions from the basic fighter maneuvering (BFM) which is chosen by the decision layer. The decision layer chooses a BFM based on the battlefield situation and monitors its execution.In the control layer, we use the fuzzy control technique, genetic algorithms as well as adaptive control technique to construct a fuzzy-adaptive controller to simulate the pilot's basic maneuvering function. This controller has small overshoot and low static error. It can be used for many types of fighter simulation models. In tactic planning layer, we consider the BFM planning as a time sequence control problem and model the BFM by the finite state machine technique. We also establish a differential game model for the three-dimensional air combat in the three-dimensional dynamic equations, which takes the fighter's aerodynamic characters, variation of speed and height, constraint of dynamic pressure, load factor and height into account. We use the shooting method to get the numerical solutions of the game. In the decision layer, we proposed a multi-knowledge bases model for the air combat decision process. In this model, all the decision knowledge bases make their decision independently. The decision layer solves the conflict depend on their priorities. We introduce a decision moderator mechanism to simulate the affect of the individual character differences (such as personality, fatigue and combat skills).We realize a fighter dogfight combat CGF system based on the proposed behavior model. It can generate human pilot like behaviors in the virtual air combat environments. It can be used in air combat training, combat simulation, and combat efficiency evaluation area.