The unconstrained optimization model for the forward positional analysis of a 6-DOF symmetrical Stewart parallel manipulator, which based on the constrained length of the bars, was presented. The standard particle swarm optimization (SPSO) has some demerits, such as relapsing into local extremum and slow convergence velocity in the late evolutionary. The improved PSO, adaptive mutation PSO (AMPSO), based on the new difference index, were proposed to overcome the demerits of the SPSO. Aimed at all forward positional solutions of parallel mechanisms were hard to obtain, stochastic algorithms were used to solve these solutions. Directed towards this weakness, the hierarchical search adaptive mutation PSO (HSAMPSO) was adopted to make the optimal problem for forward positional analysis of parallel mechanisms. Numerical results for the forward position analysis of the symmetrical Stewart parallel manipulator showed that the HSAMPSO could solve all assembly configurations, and possess the performances of rather quick convergence speed and high precision.