Cognitive radio is considered one of the effective techniques to enhance spectrum utilization [1] in modern and future wireless communication systems. In addition, the orthogonal frequency division multiple access (OFDMA) technique is a domain multiple access strategy that can allocate radio resource flexibility to multiple users [2]. Motivated by the two techniques, secondary users (SUs) in OFDMA-based cognitive radio networks (CRNs) can flexibly share the accessible licensed spectrum with primary users (PUs) if they do not introduce harmful interferences for PUs. Because of their fantastical merits, OFDMA-CRNs have been studied intensely in the past decade. However, there still exist many performance optimization problems in deploying CRNs [3]. For example, the secondary transmission often causes excessive interference to PUs. In addition, the throughput of CRN links may not be optimized for reliable high-quality communications. Furthermore, CRN links lack robustness for mobile SUs. Hence, all of the optimization problems must be solved simultaneously. It is well known that spectrum sensing [4] can solve the above problems. In previous works, SUs are assumed to be static. Hence, the instantaneous channel state information (CSI) sensed dominates the decisions of the spectrum scheduling and power allocation. However, the static assumption does not hold in mobile CRNs because of the time-variant channel. Hence, the instant sensing results are not able to guide the resource allocation (RA) independently. Moreover, when the SU is moving, the relevance of the shadow fading it suffers on the path becomes lower. This will lead to an enormous difficulty in the analysis of the channel during a continuous time slot. Therefore, the traditional sensing method and access control schemes cannot be directly applied in mobile CRNs.