A general mathematics model for cooperative task allocation of multi-UAVs with time windows constrains was proposed which incorporating task gains and execution time directly, and simplifing the model formulation and algorithm designing. By defining a suitable particle structure, an algorithm based on the principles of discrete particle swarm optimization and Guo Tao algorithm was designed. The Inver-over Operator was directed by the swarm, the local and global optimal. Variable learning selection probability is introduced into the algorithm to select the learning particles, and the Inver-Over operator was modified. Simulation verifies the proposed task planning methodology for complex missions.