For controlling the multi-robot formation system, a leader-follower separation-bearing-orientation scheme (S-BOS) is proposed and the leader-follower relationship can be represented as a formation-error kinematic system through SBOS strategy. In order to achieve the control objective, a nonlinear model predictive control (NMPC) strategy is applied to formulate the formation-error kinematic into a minimization optimization problem according to cost function. To solve this optimization problem online efficiently, a particle swarm optimization (PSO) is proposed to search for the global optimal solution as the control input. In the end of this work, simulations of the multi-robot formation are performed to verify the effectiveness of the developed strategy.