Aiming at the problem that the chest compression amount (54mm) of the THOR dummy far exceeded the high-performance limit (35mm) of C-NCAP in the Mobile Deformable Barrier Crash (MPDB) test of a certain developed vehicle model, this paper conducts an optimization study. Firstly, based on the simulation model of the qualified restraint system for benchmarking, the safety belt end plate pretensioner and the dynamic locking tongue were added, reducing the chest pressure to 46.4mm and the damage by nearly 14%. To achieve the full score goal, five parameters such as the force limiting level of the seat belt were taken as design variables, with the goal of minimizing chest compression and the constraint of head injury (HIC≤400, 3ms acceleration≤64g). The radial basis function response surface model was constructed relying on the LSOPT software, and multi-parameter collaborative optimization was carried out in combination with the genetic algorithm. Under the optimal parameter combination, the simulation verified that the chest pressure dropped to 32.73mm. Sensitivity analysis clearly indicates that the force limiting grade of the seat belt has the most significant impact on chest pressure. The final vehicle test showed that the chest pressure was 33.4mm, meeting the full score target. It confirmed that the optimization strategy was effective and reliable, providing a reference for the development of occupant protection in vehicle collisions.