Optimization problems need to be solved in many fields, and good solutions to the problems may lead to great economic benefit. With the increasing of the complexity and scale of the optimization problems, classical optimization methods that based on the strict modeling become very difficult to be carried out.The principle of artificial intelligent that based on animal's behaviors can be used for the solving of optimization problems via animals' method, then an architecture of problem solving, which we called fish school method, was defined, and a efficient intelligent optimization algorithm based on it was introduced which we called artificial fish school algorithm (AFSA).The principle and description of AFSA was described in detail. The convergence property and the influence of parameters to it was discussed; The definitions of distance, neighborhood and center were introduced to the solving of combinatorial optimization problems; For the solving of complex large scale systems, an extended artificial fish swarm algorithm based on decomposition and coordination method was described; then, some methods of improving the property of AFSA was proposed; and two applications of AFSA was discussed, one is for the parameter estimation of time varying systems, the other is for the tuning of robust PID parameters; at last, the ways where AFSA should go on was proposed.Some characters of AFSA is as follows:* The algorithm just compares the values of object functions, so it has no special requirement for the problems.* The algorithm is insensitive to the initial values, that is , either randomized or fixed initial values are all OK.* The algorithm can tolerate wide range values of parameters.* The algorithm has fast optimizing speed for the abilities of parallel processing.? The algorithm has global search abilities of optimization.The fish schools method and AFSA have different ways to the traditional method, and they have good foundations of merging with them, then we hope that they can be widely used in future.