In this presentation, we explore the application of a mirror descent approach to topology optimization. The mirror descent method, a generalized form of gradient descent, diverges from the traditional approach by replacing the quadratic Euclidean norm term with a Bregman divergence. Specifically, we leverage the Fermi-Dirac entropy as a Bregman divergence, a choice naturally derived from the bound constraint of the design variable. This divergence selection is particularly relevant as it facilitates a natural projection for the volume correction, effectively enforcing the volume constraint. The resulting projected mirror descent method seamlessly generates a sequence of bound-preserving design variables, even when employing high-order finite element approximations without additional treatment. To enhance the method's performance, we also propose an adaptive step size rule, incorporating the generalized Barzilai-Borwein method in conjunction with an Armijo-type backtracking line search algorithm. The efficacy of the proposed approach is demonstrated through several numerical examples, including applications to compliant mechanisms. Our results include comprehensive comparisons with other standard algorithms, showcasing the advantages and capabilities of the mirror descent method in the context of topology optimization.