In classical deterministic topology optimization, the effect of boundary uncertainties on the performance of the structure is not taken into account, which may lead to designs very sensitive to manufacturing errors.As a consequence, the performance of the actual structure may be far from optimal and not meet the engineering requirement.In the present paper, structural topology optimization considering boundary uncertainty during the fabrication is proposed via level set approach.In order to make the optimal design less sensitive to the boundary variations, we choose the compliance of structure enduring the worst perturbation as the objective function instead of the compliance of the original structure.In the suggested robust topology optimization approach, the boundary variations are modeled with a random normal velocity field along the boundary and a meaningful worst perturbation velocity is chosen by applying Schwarz inequality.Then the corresponding optimization problem is formulated as a worst case design problem.Numerical examples show that the robust designs achieved by the proposed approach possess a slightly worse performance in compliance but is less sensitive to the boundary uncertainty compared with minimize compliance optimization.