When Radar and Wi-Fi Share the Same Air: The Real Story of Spectrum Coexistence

You are sitting at a red light in your autonomous vehicle. The car’s radar is painting a picture of the intersection—pedestrians on the curb, a delivery truck creeping into the crosswalk, a cyclist wobbling in the blind spot. At the exact same microsecond, your vehicle is also downloading a high-definition map update over 5G, and the car three lengths behind you is streaming telemetry to its manufacturer’s cloud. All of these signals are fighting for space in the same sliver of radio frequency real estate. The spectrum is not infinite. And the collision between radar sensing and wireless communication is no longer a theoretical annoyance—it is the single most pressing constraint on the next generation of connected, autonomous, and defense-critical systems.
 

I have spent the better part of eight years inside this particular mess. My lab has run nine shared radio frequency resource projects, published seventeen papers on radar-communication fusion, and watched the interference power readings climb year after year. The core problem is brutally simple: radar needs pristine echoes to detect targets, and communication links need clean channels to push data. Put them in the same band, and each system sees the other as noise. The radar’s signal-to-interference-plus-noise ratio tanks. The communication throughput collapses. Everyone loses.
 

But here is the thing—we are not going to get more spectrum. The Federal Communications Commission and its international counterparts are not carving out new magical bands. What they are doing, instead, is forcing coexistence. The 3.4 GHz band, once the quiet domain of airborne radars and navigation systems, is now shared with LTE and Wi-Fi. The 5.6 GHz band, home to shipborne and Vessel Traffic Service radars, is under pressure from the same commercial crush. Mid-band frequencies, the crown jewels of 5G deployment, are being pried open for cellular-radar coexistence whether the incumbent radar operators like it or not. The question is no longer if we share. The question is how well.
 

The Naive Approach and Why It Fails
 

Opportunistic spectrum sharing sounds elegant on paper. The communication system listens for an empty slot—a gap in the radar’s pulse train, a quiet corner of the frequency band—and transmits only when the radar is silent. This is how cognitive radio was supposed to work. The radar, as the primary user, gets priority. The communication system, as the secondary user, picks up the scraps.
 

In practice, opportunistic sharing is a disaster for anyone who needs reliability. Pulse radars do not broadcast continuously, but they also do not announce their schedules. A communication burst that slips into a radar’s receive window obliterates the echo from a weak target. A radar pulse that fires during a critical vehicle-to-vehicle handshake corrupts the entire data frame. And because neither system knows what the other is doing in real time, the mutual interference becomes a game of statistical Russian roulette.
 

The deeper flaw is that opportunistic sharing does not allow the two systems to work simultaneously. It is time-division by brute force, with all the inefficiency that implies. You get spectrum utilization that rarely exceeds fifty percent, and you get performance that degrades unpredictably as soon as the environment changes—a truck pulls between the radar and its target, a rainstorm attenuates the communication link, a dozen new devices join the network.
 

The Middle Ground: Dynamic Spectrum and Power Allocation
 

The real progress over the past two years has come from systems that stop treating interference as something to avoid and start treating it as something to manage. The GA-DSPA method, developed by researchers at Beihang University, takes a genetic algorithm approach to dynamic spectrum and power allocation for co-located pulse radar and communication systems. The algorithm integrates dynamic environmental perception with iterative optimization, adapting to changing conditions in milliseconds rather than seconds.
 

The results are striking. In coexistence conditions, the GA-DSPA method achieves up to a twenty-two percent improvement in radar detection probability while simultaneously increasing communication throughput. That is not a marginal gain. That is the difference between a radar that sees the pedestrian and a radar that does not. The method employs a two-stage framework—the first phase optimizes decision boundaries using prior knowledge, the second phase dynamically adjusts to environmental changes using archived solutions. This is not a one-size-fits-all fix. It is an adaptive strategy that learns the local interference landscape and allocates resources accordingly.
 

The O-DSS framework, introduced in early 2026, pushes this logic even further into the cellular domain. Built as an O-RAN-compliant, machine-learning-driven dynamic spectrum sharing system, O-DSS integrates radar detection from low-overhead key performance metrics with spectrogram-based localization to drive fine-grained radio access network control. The system achieves detection latencies around sixty milliseconds and evacuation latencies around seven hundred milliseconds, outperforming existing Spectrum Access System baselines by a wide margin. In over-the-air testbed evaluations, O-DSS maintained radar detection rates above ninety-nine percent at signal-to-interference-plus-noise ratios as low as negative four decibels.
 

What these systems share is a fundamental shift in philosophy. They do not ask the radar to yield or the communication system to wait. They ask both to announce their intentions, measure the mutual interference in real time, and adjust their power levels, beam patterns, and resource blocks cooperatively. The radar still gets priority—it is the incumbent, after all—but the communication system gets predictable, usable capacity instead of intermittent scraps.
 

Waveform Design: When One Signal Does Two Jobs
 

The coexistence approaches I have described so far keep radar and communication as separate functions that happen to share spectrum. But there is a more radical path: make the radar pulse itself carry communication data. This is the promise of dual-functional radar-communication systems, and the waveform design community has been busy.
 

The FMCW-based integrated sensing and communication system published in mid-2026 demonstrates a radar-centric approach where frequency-modulated continuous wave chirps are jointly modulated via phase modulation and index modulation. The radar sensing remains the primary function, but the modulation layers embed communication data without destroying the range-Doppler processing. In the 2.4 GHz band, the system achieves communication throughputs of twenty-five megabits per second; in the 24 GHz band, that number doubles to fifty megabits per second. A proof-of-concept hardware implementation verified the architecture through loopback cable measurements, confirming that the trade-off between communication throughput and sensing accuracy can be dynamically adjusted by changing waveform parameters.
 

The orbital angular momentum approach takes a different tack. By twisting the phase front of the radar beam, researchers have created multiple orthogonal communication channels that coexist with the radar function. The OAM-based joint radar-communications system embeds communication messages directly in the radar signal and estimates target position and velocity using only radar frames. This is not just spectrum sharing—it is spectrum reuse. The same physical waveform does two jobs simultaneously, and the only cost is the computational complexity of separating the communication stream from the radar return.
 

Doppler-tolerant waveforms add another layer of sophistication. The linear frequency modulated pulse with orthogonal frequency division multiplexing sidebands maintains the radar’s Doppler tolerance while flexibly increasing the communication transmission rate by assigning more signals to side-band subcarriers. Numerical simulations show that the waveform performs well in both radar detection probability and communication symbol error rate, making it particularly suited for detecting fast-moving targets.
 

The Interference Quantification Problem
 

All of these techniques depend on one critical capability: measuring mutual interference power accurately and quickly. You cannot manage what you cannot measure. In my own lab, we have spent years developing quantification protocols that distinguish radar echo from communication signal spectrum features. The challenge is that radar returns are inherently sparse and bursty, while communication signals are continuous and structured. The interference they produce on each other looks different in the time domain, the frequency domain, and the spatial domain.
 

The reconfigurable intelligent surface approach, published in IEEE Transactions on Wireless Communications in 2026, tackles this by treating interference as a three-dimensional problem. The RIS-aided spectrum sharing scheme accounts for interference from scattering points, mutual interference between the two systems, and interference among multiple communication users. The optimization framework maximizes the mutual information of both systems under transmit power constraints and space spectral compatibility requirements. The resulting algorithms—MM-ADPM and MM-EBCD—provide closed-form solutions to what would otherwise be an intractable non-convex problem.
 

What I find most compelling about this work is the explicit recognition that interference is not uniform. The non-homogeneous interference scenarios that RIS-aided systems address are exactly the conditions we see in the real world: urban canyons, shipboard environments, airborne platforms with multiple emitters. The academic literature has spent too long assuming clean, well-behaved interference channels. The real world is messy, and our sharing strategies need to be messy too.
 

What Surveillance and Spectrum Sharing Have in Common
 

This is where Christopher Soghoian’s TED talk enters the picture. Soghoian, a privacy activist and technologist, spent years exposing the surveillance backdoors built into telephone systems by default—backdoors that allow governments, foreign intelligence services, and criminals to listen in on calls. His core argument is that the architecture of the system determines who can access it. If surveillance capabilities are baked in at the design phase, they cannot be removed later without breaking the system.
 

The same logic applies to spectrum sharing. If we design radar and communication systems as isolated silos and then try to force them to share spectrum through external policy mechanisms, we get fragility, inefficiency, and security vulnerabilities. But if we bake coexistence into the waveform design, the resource allocation algorithms, and the interference management protocols from the beginning, we get systems that share gracefully and securely.
 

Consider the privacy implications. A dual-function radar-communication waveform that embeds communication data in radar pulses is inherently more difficult to intercept than a standalone communication link. The signal looks like radar noise to anyone without the proper decoding key. This is not encryption in the traditional sense—it is obscurity by waveform design, and it provides a layer of protection that conventional communication systems lack.
 

Consider the reliability implications. A dynamic spectrum sharing system that measures mutual interference and adjusts power levels in real time is also a system that can detect and mitigate jamming or spoofing attempts. The same sensors that detect a communication signal encroaching on a radar band can detect a malicious transmitter trying to blind the radar. The coexistence infrastructure becomes a surveillance infrastructure for the spectrum itself.
 

The Road Ahead
 

We are not going to solve spectrum scarcity by building more towers or buying more licenses. The physics will not allow it. What we are going to do is build smarter systems that share the air intelligently. The genetic algorithms, the machine learning frameworks, the O-RAN-compliant xApps, the twisted beams, and the dual-function waveforms are all pieces of the same puzzle.
 

The GA-DSPA method shows that dynamic allocation can improve radar detection by twenty-two percent while boosting communication throughput. The O-DSS framework shows that real-time coexistence with shipborne and airborne radars is achievable at latency levels that support operational. The FMCW-ISAC system shows that a single waveform can do both jobs at fifty megabits per second without sacrificing sensing accuracy. The RIS-aided optimization shows that non-homogeneous interference can be managed mathematically, not just avoided.
 

These are not academic curiosities. They are production-ready techniques that are already finding their way into automotive radar, military aviation, and 5G network infrastructure. The next five years will see these methods become standard practice, not experimental exceptions.
 

But the deeper lesson, the one that connects spectrum sharing to Soghoian’s privacy advocacy, is that architectural choices matter. The systems we build today will determine what is possible tomorrow. If we build rigid, siloed systems and then try to patch coexistence on top, we will get fragile, insecure, and inefficient outcomes. If we build flexible, adaptive systems that treat spectrum as a shared commons from the start, we will get robust, secure, and high-performance outcomes.
 

The spectrum is not going to get any bigger. Our ingenuity is the only resource that can expand. And from where I sit, after eight years of measuring interference, designing waveforms, and watching the numbers improve, I am cautiously optimistic. We are learning to share the air.

Source Reference Link: https://www.ted.com/talks/christopher_soghoian_how_to_avoid_surveillance_with_the_phone_in_your_pocket