A compact genetic algorithm for the network coding based resource minimization problem

In network coding based data transmission, intermediate nodes in the network are allowed to perform mathematical operations to recombine (code) data packets received from different incoming links. Such coding operations incur additional computational overhead and consume public resources such as buffering and computational resource within the network. Therefore, the amount of coding operations is expected to be minimized so that more public resources are left for other network applications. In this paper, we investigate the newly emerged problem of minimizing the amount of coding operations required in network coding based multicast. To this end, we develop the first elitism-based compact genetic algorithm (cGA) to the problem concerned, with three extensions to improve the algorithm performance. First, we make use of an all-one vector to guide the probability vector (PV) in cGA towards feasible individuals. Second, we embed a PV restart scheme into the cGA where the PV is reset to a previously recorded value when no improvement can be obtained within a given number of consecutive generations. Third, we design a problem-specific local search operator that improves each feasible solution obtained by the cGA. Experimental results demonstrate that all the adopted improvement schemes contribute to an enhanced performance of our cGA. In addition, the proposed cGA is superior to some existing evolutionary algorithms in terms of both exploration and exploitation simultaneously in reduced computational time.