This paper present a nonlinear, model-based control of flexible link robots. The control task is formulated requiring rigid joints variables to track reference time-varying trajectory and elastic deflection to be damped. The stability and robustness properties of the control scheme are analyzed from a passive energy consideration. A direct adaptive version is also proposed. Extensive evaluation of this approach is performed using experimental validations involving a single-flexible-link and a two-flexible-link horizontal robot. Experimental results show significant performances of the controller under relatively severe working conditions: 700% payload to arm ratio and 20% elastic deflection ratio at highest acceleration stages.