The parameters characterizing the state of a vehicle occupant, which include time- invariant physical properties such as mass and stature as well as time-varying parameters such as posture and muscular bracing, influence the injury outcome in traffic collisions. To provide a feasible framework for incorporation of occupant characteristics into adaptive restraint schemes, the dissertation evaluates the sensitivity of injuries sustained in frontal collisions to four occupant state parameters—mass, stature, posture and bracing level—and describes a numerical methodology for probabilistic estimation of these occupant state parameters prior to a collision. The numerical approach included using a commercial multi-body software to develop occupant models that spanned a range of occupant state parameters representative of the real world occupant population. Coupled with a multi-body model of the vehicle interior that included a belt and airbag system, the dynamic response of the occupant models in prescribed frontal collisions was estimated through numerical simulations. Individual occupant injuries within specific body regions were predicted using established injury risk functions. A cumulative injury measure was developed to analyze the sensitivity of the injury outcome to the occupant state and other crash parameters. Further requirements of restraints to provide injury mitigation for specific occupant states were determined through optimization-based parametric simulations. For occupant state estimation during the pre-collision phase, an algorithm was developed based on information measured from conventional vehicle sensors (e.g., seat pan load cell, belt retractor) for use in probabilistic estimates of the occupant state parameters. The results from the estimation methodology characterized the probability functions for the discrete states of the occupant state parameters. An evaluation of the methodology was reported using occupant models with known state parameters and the efficiency of the algorithm was determined using standard estimator statistics. The methodology to characterize the time-invariant and time-varying occupant state parameters in the form of a probabilistic estimation model is expected to provide the necessary information for future adaptive restraint controllers to optimize and reduce injuries accounting for the stochastic nature of occupant dynamics during a crash event.