An improved shuffled frog leaping algorithm （SFLA） to solving constrained optimization problems was proposed. Combined with ε- differential evolution algorithm （ ε - DE） , the infeasible solutions with better objective function were made full use of in the evolution process. In the initial stage of evolution, the infeasible solutions with better objective function and near the boundary of the feasible region were incorporated in the population. With the evolutionary generation increasing, the decrease in the population constraint relax degree decreased the number of infeasible solutions in the population. Until the population constraint relax degree was 0, the population was entirely composed of feasible solutions. The convergence speed and accuracy had been greatly improved by the improved algorithm. The simulation results of 13 Benchmark functions proved that the algorithm had higher precision and stronger robustness, was a very effective algorithm to solve the constrained optimization problems.