Due to environmental problems, number of small vehicles with fuel efficiency increases. Since the small vehicles have small deformation space, it is difficult for them to achieve good crashworthiness at a frontal impact accident. Small deformation space usually yields high vehicle deceleration to absorb kinetic energy of the vehicle. The high vehicle deceleration may produce high occupant deceleration and lead to high occupant injury value.For example, North America, Japan and Europe specify head and chest injury value at vehicle's frontal collision. Those injury values tend to be improved if vehicle deceleration decreases. Deceleration of small vehicle with a little deformation space must be adjusted in order to prevent increase of the occupant injury value.A vehicle deceleration is expressed by 9, 18 or 36 discrete variables. A vehicle, an occupant and restraint systems such as seat belts are modeled by masses and a spring to simulate a frontal collision. With differential equations and an optimization method, the minimum value of the maximum occupant deceleration is searched when the initial vehicle velocity is 35km/h. The optimization method used is the conjugate direction method.Following results are obtained. 1Although it was high probability to be a local optimum, optimized values of vehicle and occupant decelerations were acquired. The occupant decelerations varied from 198.8 to 263.3m/s2 depending on initial values. The graph shapes were smoother than previous study. The optimized vehicle decelerations were initially high deceleration, then minus deceleration namely acceleration at mid-term, finally high deceleration.2Difference method was revealed as effective to calculate gradients of F (sensitivity). The optimizations were successfully executed with the gradients of F.3It was revealed to be effective to use sensitivity to know which part of vehicle deceleration should be modified when a way to improve a vehicle deceleration was discussed. One of examples showed an occupant deceleration reduced more than one third.