I. Introduction
To address the rising demands of data traffic, reliability, and security, the reconfigurable intelligent surface (RIS) has been proposed and thoroughly investigated over the past few years [1]. However, the widely known passive RIS-assisted systems have a double fading issue. Therefore, many researchers have combined relays with passive RIS to combat the double fading as well as harness the benefits of both. The authors of [2] utilized a decode-and-forward (DF) relay to improve the achievable rate of the system with only a weak link between the RIS and the destination. Besides, it has been demonstrated that the throughput of a RIS-relay-assisted cooperative system can be greatly improved by jointly optimizing the relay selection and the reflection coefficients of RIS using deep reinforcement learning [3]. The authors of [4] proposed both the joint and integrated RIS and relay transmission schemes, and verified that the bit error rate and achievable rate are improved significantly compared to the single RIS networks. Moreover, several buffer-aided relays were used to enhance the RIS-assisted system in [5], and the authors maximized the secrecy rate with a delay constraint by jointly optimizing the relay selection and RIS coefficients.