The size optimization of truss-plate composite structure is presented in this paper. Two improved random optimization algorithms, heuristic particle swarm optimizer (HPOS) and quick search optimizer (QGSO), are combined with finite element method to achieve the optimal design of truss-plate structure. The discrete variables are converted by the mapping function method. An example of the composite structure is given to test this two improved random optimization algorithms. The results show that the QGSO have a better convergence rate than the HPSO in the early time. They are all converged at the same fitness value and feasible on composite structure optimization.