Wireless networks have always lived on a delicate balance: more devices, higher data demands, and the constant drumbeat of interference. As researchers push toward ultra-fast 6G speeds and denser device ecosystems, the airwaves themselves become a crowded, noisy neighborhood. The result is not just slower connections, but a tangible limit on how much information we can pack into a single radio channel. In this moment of pressure and possibility, a team from the University of Surrey’s 6G Innovation Centre (6GIC) has proposed a bold new direction that could tilt the balance back toward clarity.
They offer a modulation scheme called hybrid-constellation modulation, or HCM, that blends two signaling families—QAM and ASK—into a single, superposed constellation. This isn’t just a clever trick in a lab notebook; it’s a practical design aimed at making “constructive interference” work for us, not against us. When you pair HCM with symbol-level precoding (SLP) and a reconfigurable intelligent surface (RIS) that can gently steer reflections, you get a recipe that scales better as you add more users and higher modulation orders. The work shines a light on how future networks might orchestrate signals across many antennas, surfaces, and devices with fewer computational headaches. The study is led by Yupeng Zheng and Yi Ma, with Rahim Tafazolli of 6GIC as senior author, and it foregrounds a future where radios collaborate with the environment rather than fight each other.
Hybrid modulation redefines constructive interference
At the heart of the paper is a simple, yet powerful idea: make the constellation do more work by letting one part of the symbol carry the burden of being fixed while another part is free to dance within a carefully defined constructive interference (CI) region. The authors split each transmitted symbol into a fixed part and a variable part. In the classic QAM world, most of the symbol’s geometry is locked in, and interference is either ignored or treated as noise. In HCM, the constellation is a careful blend: a portion behaves like QAM (a familiar, grid-like structure), while another portion behaves like ASK (amplitude-shift keying), with the imaginary part serving as a playground for CI exploitation. This clever split effectively expands the CI region(s) that the system can exploit, making it easier for the transmitter to align multiple users’ signals so they reinforce rather than undermine one another.
To visualize it, picture a city built on two architectural styles side by side—rectangular blocks (QAM) and slender towers (ASK). The real dimension (the real axis) locks in some signals, while the imaginary dimension becomes a flexible space where interference can be shaped and steered in helpful directions. The authors illustrate that 16-HCM uses an 8-ASK plus 8-QAM mix, while 64-HCM uses 16-ASK plus 48-QAM. The upshot is a constellation that is not only denser on paper but more forgiving in practice because the interference that would usually cause trouble can now be harnessed as a signal-boosting feature. This isn’t a theoretical curiosity; the simulations show meaningful gains in the error rate, especially as the modulation order climbs.
Why this could reshape 6G and beyond
The practical punch line is about scalability and robustness. RIS is already touted as a way to extend wireless coverage by smartly reflecting signals through a programmable sheet of tiny mirrors. But tying RIS to symbol-level precoding introduces two thorny issues: how to optimize the many possible symbol interactions without exploding the computation, and how to do it when the RIS can only implement a finite set of phase shifts. The Surrey team tackles both head-on. They propose a two-stage approach: first, optimize the RIS phase shifts using a low-complexity method that works with a discrete set of phase angles, and then optimize the transmit vector for the chosen RIS configuration. This separation avoids the astronomical combinatorial explosion that would occur if you tried to search all possible symbol configurations jointly.
The second stage leans on a mathematical technique called widely linear (WL) processing, which helps separate the fixed and variable parts of the symbols and recast the problem in a form that a solver can handle. They also introduce a practical, near-optimal shortcut for the SLP step that avoids a brute-force search over all possible sign patterns. The result is a method that scales better as you add more users or raise the modulation order, while still delivering the kind of gains that make SLP compelling in the first place. In numerical terms, the authors report up to 1.5 dB better symbol error rate (SER) performance for 16-ary schemes and about 1 dB for 64-ary schemes, compared with conventional QAM-based SLP. Crucially, these gains persist even when the RIS elements have only a few discrete phase levels (1-bit or 2-bit phase shifters), a realism constraint that often sinks theoretical promises.
From ideas to real systems with discrete RIS
The leap from elegant equations to a deployable system is nontrivial. Real-world RIS hardware frequently uses discrete phase shifts rather than a continuous range, and that discreteness can threaten performance. The Surrey paper is careful about this gap. Their results show that even with coarse RIS phase resolution, the HCM-SLP approach retains substantial gains over traditional QAM-SLP. In some cases, a 1-bit RIS (two phase choices) performs nearly as well as a higher-resolution RIS, because the extended CI regions from HCM compensate for the reduced phase fidelity. This is the kind of robustness that makes a research advance feel closer to a practical breakthrough than a purely theoretical one.
The two-stage framework is also a win for engineering teams worried about computational budgets. Rather than enumerating every possible symbol combination—an approach that becomes untenable as K (the number of users) grows—the method incrementally refines the RIS configuration and then finishes with a tractable optimization that leverages the special structure of HCM’s fixed-and-variable symbol parts. In other words, the method is designed with the rough realities of hardware and software in mind, not just the tidy world of mathematical proofs.
Bottom line: this work from the University of Surrey’s 6GIC shows a path toward higher data rates and lower error rates without demanding unrealistically perfect hardware. It reframes how symbols can be designed and how reflective surfaces can be tuned so the “air” becomes a more cooperative partner in delivering data—especially as networks grow more crowded and users demand more reliable connections.
Beyond the immediate technical novelty, the research is a reminder of how progress often hinges on reimagining the rules of engagement. The idea that interference can be constructive, that a signal constellation can host two signaling philosophies at once, and that a distributed surface in the environment can be tuned to amplify those gains together with the transmitter—these are the kind of shifts that could ripple across the future of wireless, from how we deploy networks in dense urban canyons to how devices stay connected in crowded venues.
As with many foundational ideas in wireless communication, the real test will be translation to hardware, firmware, and standardization. The authors acknowledge that their work is framed within specific assumptions about channel models and coherence times, and that practical deployments will require robust calibration, testing in real-world channels, and continued refinement of the optimization algorithms. Still, the study is an important step in answering a central question of modern wireless design: how can we make the environment work for us, not against us, as we push toward increasingly demanding performance targets? The University of Surrey’s team has given industry and academia a tangible blueprint for doing exactly that, with a thoughtful blend of mathematical insight and engineering pragmatism.
In the end, the researchers’ success is not just about packing more bits per hertz; it’s about teaching our networks to listen—and to adapt—so the airwaves feel less like a battlefield and more like a well-rehearsed ensemble. The lead authors, Yupeng Zheng and Yi Ma, along with senior author Rahim Tafazolli, have sketched a future where hybrid signal design and friendly surfaces join forces to extend coverage, improve reliability, and keep up with our ever-growing appetite for data. It’s a future that may still be a few years away, but the map is now clearer—and the music might just be tuning up.
