In this paper, the proposed method adopts a novel dynamic genetic algorithm to solve starvation in unicast routing problem for WMN in the network layer and MAC layer. To avoid the slow convergence of the GA and utilize the fast convergence of GSA, a novel hybrid algorithm to localize the solution space is proposed. Simulations are conducted for multimedia traffic. The results obtained are compared with priority-based starvation avoidance method and GA-based optimization without hybridization of GSA. Performance is compared in terms of parameters such as throughput, end-to-end delay and number of cached replies used.