SUMMARYTo solve the problem of cognitive link decision for OFDM system in a time-varying environment, an improved genetic algorithm-based decision-making module is proposed. At first, the decision-making module collects the environmental information and QoS requirements of the radio and then produces link configuration for the radio on the basis of the improved genetic algorithm. New crossover and mutation strategies are introduced to accelerate convergence rate of the genetic algorithm. A novel population initialization method is also proposed to reduce the time of achieving the optimal decision by the decision-making module when the radio is operated in a slow changeable environment. Simulations and discussions demonstrate the effectiveness of the decision-making module, and this research work exhibits its importance of enhancing the decision speed that is of crucial demand in cognitive decision-making applications.