Modern aircrafts are equipped with more control effectors to enhance the reliability and survivability. Control allocation is used to deliver the virtual control moments to redundant control effectors. Due to the physical constrains and aerodynamic considerations, each control effector has some constrains and the control allocation becomes more complicated. A new control allocation algorithm based on weighted pseudo-inverse is proposed in this paper. Weighted pseudo-inverse is a control allocation method with easy structure and high computational efficiency, but it could not give out the optimal result. Genetic algorithm is a randomized search algorithm. In this paper, it will be used to optimize the weighting matrix to achieve maximum attainable moment set. To enhance the performance of the algorithm, the space of the desired moment will be divided into several parts and each part will get one optimal weighting matrix obtained by genetic algorithm. Multiple weighting matrices could guarantee better performance in each part than one weighting matrix. Direct allocation is used to be contrast to demonstrate the effectiveness of the algorithm. Simulation results show that the new proposed algorithm could achieve much better performance than the traditional weighted pseudo-inverse method.