Unmanned aerial vehicles (UAVs) are powerful Internet-of-Things components to provide sensing and communications in the air due to their advantages in mobility and flexibility. As aerial users, UAVs are envisioned to support various sensing applications in the next generation cellular systems, which have been studied by the Third Generation Partnership Project (3GPP). However, the Quality-of-Services (QoS) of the cellular link between the UAV and the base station may not be guaranteed when UAVs are at the cell edge or experiencing deep fading. In this article, we first introduce the non-cooperative cellular Internet of UAVs. Then we propose a cooperative sense-and-send protocol, in which a UAV can upload sensory data with the help of a UAV relay, to provide a better communication QoS for the sensing tasks. Key techniques including trajectory design and radio resource management that support the cooperative cellular Internet of UAVs are presented in detail. Finally, the extended cooperative cellular Internet of UAVs is discussed for QoS improvement with some open issues, such as massive multiple-input multiple-output systems, millimeter-wave, and cognitive communications. arXiv:1904.01257v2 [cs.IT] 3 Apr 2019 2 Fig. 1. Illustration for UAV sensing applications.Recently, the Third Generation Partnership Project (3GPP) has approved the study item on enhanced support to seamlessly integrate UAVs into future cellular networks [4]. Notably, it has produced accurate modeling of UAV-to-ground channels, and defined various UAV scenarios together with their respective features. Therefore, the terrestrial cellular networks, e.g., LTE and 5G, are considered as a promising enabler for UAV sensing applications, which we refer to as cellular Internet of UAVs. In the cellular Internet of UAVs, the sensory data needs to be transmitted to the base station (BS) for timely processing. However, when UAVs are located at the cell edge or experiencing deep fading, the quality of the communication between the UAV and the BS may not be satisfactory, specially for emergency search-and-rescue and surveillance applications where the sensing task is far away from the BS. In the non-cooperative cellular Internet of UAVs, the UAV should fly to a communication point first to satisfy the QoS for the communication link, which will bring extra delay.To tackle this challenge, we propose the UAVs to transmit via a UAV relay for providing a better QoS, as shown in Fig. 2. To support the cooperative cellular Internet of UAVs, the system provides two types of communications, i.e., UAV-to-network (U2N) and UAV-to-UAV (U2U).