This paper proposes a new robust adaptive law for adaptive control of vehicle active suspensions with unknown dynamics (e.g. non-linear springs and piece-wise dampers), where precise estimation of essential vehicle parameters (e.g. mass of vehicle body, mass moment of inertia for the pitch motions) may be achieved. This adaptive law is designed by introducing a novel leakage term with the parameter estimation error, such that exponential convergence of both the tracking error and parameter estimation error may be proved simultaneously. Appropriate comparisons with several traditional adaptive laws (e.g. gradient and -modification method) concerning the convergence and robustness are presented. The mitigation of the vertical and pitch displacements can be achieved with the proposed control to improve the ride comfort. The suspension space limitation and the tyre road holding are also studied. A dynamic simulator consisting of commercial vehicle simulation software Carsim® and Matlab® is built to validate the efficacy of the proposed control scheme and to illustrate the improved estimation performance with the new adaptive law.