On-line Sate-of-Charge (SOC) estimation in pure electric or hybrid-electric (HEV) vehicles is a challenging problem, due to the very dynamic nature of the current/voltage history under actual driving conditions. However, on-line, reliable SOC estimation is critical in these applications, particularly in charge-sustaining HEV, where the battery capacity is relatively small and where the energy management strategy needs an accurate SOC estimation in order to optimize the power split between the ICE and EM. The research described in this paper focuses on two aspects of the same problem. The first aspect is the development, calibration and validation of a dynamic battery model which represents the essential dynamic behavior of the battery pack HEV application. The second part of this paper is the development, implementation and validation of an appropriate in-vehicle SOC estimator for use with our supervisory HEV controller. While the implementation approaches for theses two facets of the same problem are significantly different due to the real-time computational requirements, they represent the same battery dynamics. This algorithm has been applied to a real charge-sustaining HEV vehicle and the experimental results are presented.