The polyphase code signals win popularity and attention for its big output main -side ratio and high security, the cost function of the MIMO radar polyphase code optimization is set up based on auto-correlation sidelobe peak, cross correlation peak, the total auto-correlation sidelobe energy and the total cross correlation energy. Then, based on the idea of the multi-swarm cooperation and the high level swarm's high effective decision, the normal particle is utilized to search in optimize space, and the high level particles lead the swarm convergence to the global best solution more quickly. As a result, a new algorithm called Multi-swarm delaminated cooperation Genetic Algorithm is proposed and applied to the MIMO radar polyphase code design, which enables the MIMO radar signals to own high resolutions in both distance and velocity measurements, the computational simulation results show that the proposed algorithm is effective and reliable.