In this paper, we study dynamical traffic light strategies in the Biham-Middleton-Levine traffic flow model. The strategies use local vehicular information to control urban traffic, which take into account the interaction of vehicles traveling in different directions via considering their dynamical spatial configuration. Simulations find out two strategies, in which local information at nearby sites is used. The two strategies perform much better than the alternating strategy. Under these two strategies, vehicles can self-organize into a new intermediate state with band structure. The analytical solutions of velocity of this state have been presented, which are in good agreement with simulations.