Variable environments in ad hoc networks require a joint control of physical (PHY) and medium access control (MAC) layers resources in order to optimize performance. In this paper, we propose a framework to perform such cross-layer control and optimization. The PHY layer and cross-layer engine estimate and predict the channel variations to select the users that will meet the signal-to-interference-noise ratio (SINR) requirement in the next time slot, for MAC layer optimization. We consider high capacity code division multiple access (CDMA) ad hoc networks working at fixed quality of service (QoS) requirement where nodes are equipped with matched filter receivers.