As humanoid robots shift from research lab to human living environment, biped robots play a more important role in recent research. Peoples are attracted by bip ed robots for the difficulties in gait control since their dynamic equations are high-order highly-coupled and nonlinear. This paper presented the basic concept s in biped gait and major research methods in the filed of biped gait optimizati on and control. Characters of three major models including inverted pendulum, pa ssive gait model and mass-spring model are thoroughly discussed. Advantages and disadvantages of all these three models are exhibited by comparison. Followed wi th it, the two kinds of usually used constraints, stability criterion and energy constraint are introduced with abundant examples. For stability criterion, zero moment point (ZMP) criterion as the most useful constraints in biped gait gener ation research is emphasized. Finally three intelligent control methods (neuron network, fuzzy logic and genetic algorithm) with their combination methods (fuzz y neuron network, genetic neuron network and so on) were introduced at the end o f this paper.