In (Shtaiwi et al., 2021), a large RIS is used in a multiple-input multiple-output (MIMO) RCC system to achieve a high beamforming gain and mitigate MI between communication and RADAR system, thereby improving the sum rate of the communication system subjected to the radar performance constraints. The transmit power of both the radar transmitter and the BS is kept within their respective budgets. The optimization problem of the BS transmit beamformer and the RIS phase shifters is solved using a local search approach. Incorporating RIS in NLOS systems establishes a communication link between users and the BS, and it is confirmed that using more RIS elements improves beamforming gain, enhances interference mitigation, and leads to a monotonically increasing sum-rate in the communication system. An RIS-aided multi-user MIMO RCC system is illustrated in Figure 1A.

A double-RIS RCC system is considered in (He et al., 2022) to improve the communication performance and reduce MI while guaranteeing the radar detection performance for each detection direction. To achieve this objective, a double-loop penalty dual decomposition-based algorithm is employed. Two scenarios are considered in which the radar transmit power is respectively high and low. In the former, a closed-form optimal solution for joint beamforming is derived, while in the latter, the MI is minimized using the block coordinate descent method. It is shown that introducing a single 100-element RIS to an RCC system improves the output communication signal-to-interference-plus-noise ratio (SINR) by 6 dB. On the other hand, a double-RIS-assisted system further suppresses interference and enhances the output communication SINR by 11 dB, highlighting the effectiveness of deploying two RISs.

Studies conducted on DFRC systems that utilize RIS are concentrated on either optimizing communication performance while upholding a certain level of sensing performance, or optimizing the sensing performance while preserving a certain communication rate. Figures 1B, C respectively depict the scenarios where RIS assists either the communication or the sensing function. In Figure 1D, the targets are detected through virtual LOS connections from the BS to the targets via the RIS.

In (Liu et al., 2022b), a distributed RIS system consisting of both passive and semi-passive RISs conducts location sensing and data transmission simultaneously for a single user scenario. A semi-passive RIS contains some active sensors that can be used to estimate the CSI associated with the RIS. Unlike other ISAC systems which aim to perform separate sensing and communication functions, the purpose of sensing here is to assist the communication function in the sense that the estimated CU location facilitates the optimization of the RIS beamformer. Each transmission period is divided into two-time blocks. During each time block, passive RIS elements reflect the uplink transmission signals from the CU to the BS while the semi-passive RIS elements estimate the CU’s location in a sensing mode. This is achieved by using the total least squares ESPRIT method and pairing the effective two-dimensional DOAs using the MUSIC algorithm. The estimated location from the first time block is then used for beamforming design at the BS in the second time block. By implementing this location sensing scheme, it is possible to achieve millimeter-level positioning accuracy. Additionally, the proposed beamforming design enables near-optimal bit rates at a single CU, even when only a small number of RIS elements are used.

In (Asif Haider et al., 2023b), a semi-passive RIS-aided ISAC system is considered where a small number of active RIS elements are sparsely placed in an L-shape structure, as shown in Figure 1F. With the large array aperture, this structure improves channel estimation performance and reduces the required length of the pilot overheads. The CU and target channels are first estimated using a two-dimensional interpolated covariance matrix of the array data signal and the MUSIC algorithm. Joint optimization of the BS beamformer and RIS phase shifters maximizes the beamforming gain towards the direction of desired sensing and ensures minimum output SNR at the CU. The study provides insights into the design for different arrangements of the active sparse elements for performance improvement without increasing hardware costs.

A deep learning-based framework is considered for the estimation of the sensing and communication channels in an RIS-aided ISAC system (Liu et al., 2022c). Two deep neural networks (DNN) architectures are proposed where the first one is located at the ISAC BS to estimate the sensing channel, whereas the second one is assigned to each downlink CU equipment to estimate its communication CSI. The BS is operated in a full-duplex mode. The self-interference due to full-duplex operation is pre-estimated and compensated before estimating the BS-target-BS channel. The deep learning framework contains three hidden layers with two convolutional layers and a flattened layer which are used for the estimation of the communication CSI, whereas two feed-forward layers with the tanh activation function are used for sensing channel estimation. By training the DNN using multiple SNR conditions, this method can be applied to a broad range of SNR regions. It achieves better sensing and communication channel estimation performance compared to the least squares estimator benchmark scheme.

RIS beamformers often exploit discrete phase shifters to facilitate low-cost hardware implementations compared to continuous phase shifter counterparts. Discrete phase shifters use a limited number of phase levels, such as those based on PIN diodes or varactor diodes (Wu and Zhang, 2019). It is shown, however, that there is a trade-off between the discrete quantization levels and the achievable sum rate as well as sensing performance (Wang et al., 2021; Wei et al., 2022b; He et al., 2022; Zhu et al., 2022).

Wideband signals support the transmission of a high data rate and provide higher sensing resolutions. An RIS-aided wideband DFRC system with multi-carrier signaling is considered in (Wei et al., 2022b). Because the direction of the beam can change significantly over the operating frequency range, the beam squint problem becomes more severe as the signal bandwidth increases. When this problem is not properly addressed, the beam may no longer be aligned with the intended target, thereby leading to degraded performance. The quantized phase-shifter model utilizes frequency-dependent transmit beamforming to overcome the beam-squint effect in a wideband DFRC system. By jointly designing the radar receive filter, transmit precoding matrix, and discrete phase shifters, the radar output SINR is maximized while the required level of communication SINR is maintained for single-antenna CUs. The resulting optimization problem is solved using the alternating maximization method.

In ISAC systems, the inclusion of sensing functionality presents a significant challenge as it limits the available degree of accuracy for waveform design, leading to increased MUI and reduced communication performance. In (Wang et al., 2021), RIS is explored to mitigate MUI, with joint optimization of beamforming at the BS and discrete RIS phase shifters to meet the Cramer-Rao bound (CRB) requirement for multi-target DOA estimation. To achieve this objective, the exact penalty method and manifold optimization approach are used for constant-modulus waveform optimization at the ISAC BS, and manifold optimization and successive optimization methods are used to optimize the discrete RIS phase shifters. The proposed algorithm yielded a significantly increased multi-user sum rate compared to pure ISAC systems without RIS.

As shown in Figure 1G, multiple RISs can be utilized in ISAC systems to further enhance object tracking performance, maximize communication bit rate, mitigate MI, and increase user SINR (Wei et al., 2022a; He et al., 2022). In (He et al., 2022), the use of multiple RISs in an RCC system improves the MI suppression performance compared to a single-RIS scenario. The study in (Wei et al., 2022a) focuses on the joint design of frequency-dependent beamforming and phase shifters for a wideband orthogonal frequency division multiplexing (OFDM) transmit signaling scheme, aiming to maximize the average radar output SINR and the minimal communication output SINR among all users using an alternating maximization method. The non-convex problem comprises a maximin objective function with a difference of convex constraint, which is solved through the alternating direction method of multipliers (ADMM) and Dinkelbach’s methods. It is shown that, by utilizing two RISs, the proposed method achieves a 3.3 dB improvement of the radar output SINR improvement and a 0.9 dB enhancement of the minimal communication output SINR compared to the single-RIS counterpart.

In (Shao et al., 2022), the BS communicates with multiple devices in a full-duplex mode while simultaneously sensing their positions. As shown in Figure 1H, each device is equipped with an RIS to enhance the reflected echoes and passively transfer information to the BS through the reflection modulation. To address the challenges of joint localization and information retrieval at the BS, a grid-based parametric model is constructed and the problem is formulated as a compressive sensing problem. The study focuses on designing the BS receiver to jointly locate the positions of the devices and retrieve the information, which is transmitted from the CUs and transferred by the RISs, through the exploitation of the sparsity in the considered parameter grid. The results confirm that the proposed algorithms outperform commonly used compressive sensing algorithms, such as orthogonal matching pursuit (OMP) and sparse Bayesian learning (SBL), in terms of the achieved minimal error in the recovered transmit information, particularly in multi-user off-grid scenarios.

The presence of clutter can have a significant impact on the performance of target detection in ISAC systems. In (Liu et al., 2022d), the problem of target detection is considered in the presence of strong clutter while ensuring the quality of multi-user communications. The space-time adaptive processing technique is used in the BS to achieve effective target detection in the presence of strong clutter that spreads out over large ranges and angular regions. By optimizing the nonlinear spatio-temporal transmit waveform and receive filter, this method makes use of all available degrees of freedom in both the spatial and temporal domains to improve the target detection performance.

The issue of eavesdropping in a DFRC system, as illustrated in Figure 1I, is considered in (Mishra et al., 2022). In this system, the radar target acts as an eavesdropper that receives a high-energy signal, leading to additional security challenges. A multi-antenna DFRC system is considered in the presence of multiple eavesdroppers, and a physical-layer approach is designed to maximize the secrecy rate while meeting the total transmit power constraint and maintaining a required output SINR for each radar target. The multi-eavesdropper multi-cast multi-antenna DFRC system is optimized to maximize the secrecy rate. The RIS is used to enhance the quality of signal transmission to users, which indirectly allows additional degrees of freedom for the injection of artificial noise to distort signals arriving at eavesdroppers.