In traditional wireless network design, the physical propagation environment is treated as an uncontrollable, random black box. Engineers design increasingly sophisticated transmitters and receivers to adapt to fading, multipath propagation, and shadowing, but the wireless channel itself remains entirely passive.
Reconfigurable Intelligent Surfaces (RIS)—also referred to as Intelligent Reflecting Surfaces (IRS)—fundamentally break this paradigm. An RIS is an artificial planar surface embedded with hundreds of low-cost, passive software-controlled reflecting elements. By dynamically shifting the phase of incident electromagnetic waves, an RIS can alter the wavefront to achieve constructive or destructive interference at specific receiver locations. As outlined by Xiaohu You et al. in Reconfigurable Intelligent Surfaces for 6G Systems: Principles, Applications, and Research Directions, this capability effectively transforms the wireless channel from a fixed environmental constraint into a programmable software variable, capable of doubling aggregate data rates in dense multi-cell environments.
The Synergy of RIS and Power-Domain NOMA
The true power of environment programming becomes evident when paired with Non-Orthogonal Multiple Access (NOMA). Unlike traditional Orthogonal Multiple Access (OMA) schemes that segregate users into distinct time slots or frequency bands, NOMA allows multiple users to share the exact same time-frequency resources by separating them in the power domain.
At the transmitter, signals for multiple users are superimposed with different power levels. At the receiver, Successive Interference Cancellation (SIC) is used to decode and strip away interfering signals sequentially. Crucially, NOMA relies on a distinct disparity between user channel gains to function efficiently. If a "strong" user (near the base station) and a "weak" user (far away at the cell edge) have highly contrasting channel qualities, the receiver can easily distinguish and isolate their signals. However, if their natural channel gains happen to be similar, SIC becomes highly error-prone, and NOMA's performance degrades.
An RIS solves this core limitation by intentionally engineering the channel gain disparity. Instead of accepting whatever channel states nature provides, the RIS adjusts its phase shifts to constructively reinforce the strong user's signal while simultaneously creating destructive interference or mitigating path loss for the weak user.
[Base Station] ----(Superimposed Signal)----> [ RIS (Phase Optimization) ]
/ \
(Constructive) / \ (Destructive/Balanced)
v v
[Strong User] [Weak User]
(High Channel Gain) (Low Channel Gain)
This deliberate channel engineering creates an ideal environment for SIC. Research by Xidong Mu et al. in Exploiting Intelligent Reflecting Surfaces in NOMA Networks: Joint Beamforming Optimization and Tianwei Hou et al. in Reconfigurable Intelligent Surface Aided NOMA Networks demonstrates that jointly optimizing the base station's active beamforming and the RIS's passive phase shifts yields massive spectrum and energy efficiency gains over conventional OMA systems.
Furthermore, performance analyses by Kwok Hung Li et al. in Downlink and Uplink Intelligent Reflecting Surface Aided Networks: NOMA and OMA and Xinwei Yue et al. in Performance Analysis of Intelligent Reflecting Surface Assisted NOMA Networks show that this combination significantly boosts diversity order and ergodic rates, even when constrained by practical limitations like coarse 1-bit phase quantization or imperfect SIC. Using stochastic geometry, Chao Zhang 0048 et al. in Reconfigurable Intelligent Surfaces Aided Multi-Cell NOMA Networks: A Stochastic Geometry Model proved that this synergy is particularly potent for cell-edge users, rescuing them from deep fading and severe co-channel interference.
Overcoming the Double Path-Loss Dilemma
Despite its advantages, a purely passive RIS faces a fundamental physical hurdle: double path loss. Because a passive surface does not generate or amplify power, the signal traveling through the RIS link experiences attenuation twice—once from the transmitter to the surface ($d_1$), and once from the surface to the receiver ($d_2$). The resulting path loss is multiplicative ($\propto 1 / (d_1^2 \cdot d_2^2)$), making the reflected signal exceptionally weak compared to the direct line-of-sight path.
To overcome this severe attenuation, researchers have introduced two architectural solutions that combine active signal processing with passive reflection.
1. Active RIS Elements
Instead of merely altering the phase of the incident wave, an active RIS incorporates low-power reflection amplifiers into its elements. As shown by Feng Shu 0002 et al. in Three High-Rate Beamforming Methods for Active IRS-Aided Wireless Network, designing optimal reflection coefficients with active amplification provides a crucial power boost that directly counteracts deep fading, delivering significantly higher data rates than purely passive surfaces.
2. Hybrid Relay-Reflector Systems
A second paradigm pairs passive surfaces with active Decode-and-Forward (DF) relays. In this configuration, the active relay breaks the multiplicative path-loss chain by receiving, decoding, regenerating, and re-transmitting the signal. Concurrently, the passive RIS shapes the spatial propagation environment to maximize beamforming efficiency and eliminate inter-user interference.
The authors of Phase Optimization for Massive IRS-aided Two-way Relay Network (Siling Feng et al.) and Coexisting Passive RIS and Active Relay-Assisted NOMA Systems (Ao Huang et al.) demonstrate that this active-passive coexistence creates a powerful division of labor. The active relay injects power and regenerates the signal, while the passive RIS handles spatial programming. This hybrid approach yields superior sum rates and user fairness compared to networks relying on either technology alone.
Scaling and Real-Time Optimization
The performance gains of an RIS scale directly with the number of reflecting elements, as a larger array provides more degrees of freedom to focus electromagnetic energy. Xu Zhu et al. in Robust Optimization for IRS-Assisted SAGIN Under Channel Uncertainty demonstrate that scaling up the number of elements provides the necessary redundancy to maintain a robust signal-to-interference-plus-noise ratio (SINR) even under severe channel state uncertainty. Similarly, Yu Zhang 0015 et al. in Beamforming and fronthaul compression design for intelligent reflecting surface aided cloud radio access networks show that integrating RIS phase control with uplink fronthaul compression significantly mitigates quantization noise in Cloud Radio Access Networks (C-RANs).
However, as arrays scale to hundreds of elements, calculating the optimal phase shift configuration in real time becomes a massive computational challenge. When combined with the need to optimize base station power allocation and dynamic user trajectories, traditional iterative optimization algorithms become too slow for practical deployment.
To bypass this computational bottleneck, researchers are leveraging machine learning to predict optimal configurations instantly:
- Meta-Learning: Yixuan Zou et al. in Meta-learning for RIS-assisted NOMA Networks developed a meta-learning framework that learns the underlying structure of the joint phase-shift and power-allocation problem. This allows the network to adapt to new channel conditions with a fraction of the training iterations required by standard deep learning models.
- Deep Reinforcement Learning: In highly dynamic environments, such as multi-robot coordination, Xinyu Gao et al. in Intelligent Trajectory Design for RIS-NOMA Aided Multi-Robot Communications utilize deep reinforcement learning to jointly track robot movements, plan physical trajectories, and update RIS phase configurations on the fly, ensuring continuous, high-rate connectivity.
Key Foundational Work
Reconfigurable Intelligent Surface Aided NOMA Networks (2020) — contributed 2 concept elements to this direction.
Meta-learning for RIS-assisted NOMA Networks (2021) — contributed 1 concept elements to this direction.
Downlink and Uplink Intelligent Reflecting Surface Aided Networks: NOMA and OMA (2021) — contributed 1 concept elements to this direction.