The Big Bang-Big Crunch (BB−BC) algorithm is a recently developed metaheuristic optimization method that mimics the process of evolution of the universe. The inherent simplicity of BB−BC is very attractive for structural optimization experts but the huge computational cost entailed by the optimization process is a serious obstacle to the large-scale diffusion of BB−BC. To overcome this problem, at least in the case of sizing optimization problems of truss structures, the paper presents an improved BB−BC formulation where each new trial design is always forced to lie on a descent direction. The efficiency of the new BB−BC algorithm is demonstrated by a trade study carried out on four classical weight minimization problems of truss structures that include various amounts of non-convexity in the design space