We extend our previous work on optimal dynamic resource allocation in wireless environments to incorporate prediction of the frequency-selective OFDM channel. We briefly summarize our previous work and its exploitation of convexity for the resource allocation problem in point-to-point digital wireless communication links. We introduce channel prediction to overcome latency associated with symbol recovery, channel estimation, and channel-state feedback, which previously restricted resource allocation algorithms to implementation in slowly-fading channel environments. The resource-allocation framework is augmented with channel prediction functionality, and we demonstrate its use with a channel model exhibiting frequency-selective fading with a limited time autocorrelation. Results are presented, illustrating successful implementation, and we conclude with an outline of the course of future investigation to make channel-prediction-based resource allocation a viable technique in practical OFDM systems.