Aimed at the shortcoming of easy plunging into local optimum of krill herd algorithm (KH) in solving high dimensional complex optimization problems, an improved krill herd algorithm based on mutualism and random disturbance is proposed.First, the algorithm puts the strategy of mutualism into the standard KH, which enhances the exchange of particles' information, improves the efficiency of information transfer effectively and increases krill individual's ability of survival.Second, in order to avoid the premature convergence, this algorithm adopts the evolution of the survival of fittest in the natural selection mechanism to improve the quality of the individuals.Finally, the experiments simulating 12 benchmark standard test functions show that the algorithm has good effect in optimization capability and accuracy of solution, and is surperior to or highly competitive with other two kinds of algorithm.