Radio resource management (RRM), which aims to satisfy the requirements of both mobile users and service providers, can be seen as one of the typical issues of cyber-physical-social system since the social factors, that is, the requirements and priorities of users are extremely important in heterogeneous networks. In this paper, we propose a novel resource allocation and access control mechanism based on parallel network architecture, which provides a high-bandwidth connectivity with guaranteed quality of service (QoS) for mobile users in a seamless manner. In this mechanism, multiple users are classified into several types according to their social property such as priorities and bandwidth requirements of different users. Compared with the general received signal strength (RSS)-based method, the proposed user priority (UP)-based method achieves three main advantages as follows: 1) it further balances the load of base stations (BSs) when the resource is sufficient; 2) it provides a mechanism called high priority users higher QoS when the network is heavily loaded compared to the RSS-based method; and 3) it hands over a few users from a heavily loaded BS to a lightly loaded one to allow more users to access this network. The simulation results confirm the advantages of the proposed UP-based mechanism and show that the simulation results of the Q-learning method are consistent with its theoretical analysis.