Abstract: For particle swarm algorithm not being applied to the rapid evolution of virtual biological cluster in short range, a particle swarm optimization method was provided that oriented to interactive intelligent fish with weight dynamic constrained. This method let particle swarm through the state of the particle separation, and dynamic constraint particle swarm. This method used the concept of "convergence coefficient manager" to retain the differential movement between the particles. On this basis, setting the evaluation function and using the dynamic constraint weights, the fast particle swarm was completed, applying to virtual biological cluster. The experiments results show the best weights dynamic constraint effect in large-scale virtual biological cluster and intelligent fish pattern movement, and this method is more effect than common particle swarm optimization algorithm, and accelerates significantly. And this method has been used in the development of virtual aquarium system with stable and reliable.

Key words: artificial intelligence fish, dynamic constraint weights, particle swarm, evaluation function