Beamforming and Filter Design for Bistatic ISAC under Known and Unknown Transmit Symbols

In the crowded landscape of modern wireless networks, every bit of radio spectrum is a precious resource. For decades, engineers have treated two critical functions—sending data to phones and detecting objects like cars or drones—as separate operations, each requiring its own hardware and its own slice of the radio frequency spectrum. This separation is becoming inefficient as the demand for both faster internet and more precise sensing grows. A new approach, known as integrated sensing and communications, seeks to merge these two worlds. Instead of using different signals for talking and listening, this technology uses a single, shared signal to do both simultaneously. The challenge lies in designing this shared signal so that it carries information clearly to a user's device while also bouncing off distant objects with enough clarity to be detected by a radar system.

The researchers behind this study focused on a specific and practical arrangement called a bistatic system. In a traditional radar setup, the transmitter and receiver sit in the same location, which creates a difficult problem: the powerful signal sent out can overwhelm the receiver, much like trying to hear a whisper while standing next to a roaring jet engine. To avoid this, a bistatic system places the transmitter and the receiver far apart. The transmitter, usually a cellular base station, sends out a signal that serves two purposes: it delivers data to multiple users and it illuminates targets in the environment. The receiver, located at a different site, picks up the faint echoes bouncing off those targets. The goal of this work was to figure out how to shape the signal sent from the base station and how to process the signal at the receiver so that both the communication quality and the radar detection performance are as good as possible, even when the system is under strict power limits.

The team tackled a complex design problem by considering two very different real-world scenarios regarding what the radar receiver knows about the signal it is listening for. In the first scenario, the receiver has full knowledge of the data symbols being transmitted; it knows exactly what message was sent. In the second, more difficult scenario, the receiver has no idea what the symbols are, perhaps due to privacy restrictions or a lack of coordination between the systems. The researchers developed mathematical methods to optimize the signal shaping and the filtering process for both cases. Their aim was to maximize the clarity of the radar echoes, specifically the signal-to-interference-plus-noise ratio, which is a measure of how distinct a target's echo is from background noise and other interfering signals, while ensuring that every communication user still receives a clear connection.

When the symbols are known, the researchers found a way to calculate the best possible filter for the receiver in a direct, closed-form solution. They then used an iterative process to refine the signal sent from the transmitter, gradually improving the radar's ability to see targets without dropping the quality of the communication links. When the symbols are unknown, the problem becomes significantly harder because the receiver cannot simply subtract out the known signal to find the echo. In this case, the team employed a step-by-step approach, alternating between optimizing the transmitter's signal and the receiver's filter. They would fix one part, solve for the best version of the other, and then switch back, repeating the cycle until the system settled into a stable, high-performing state.

The results of their simulations, conducted with a base station and a receiver each equipped with sixteen antennas, revealed a clear distinction between the two scenarios depending on how much time the system had to observe the targets. When the system only looked at a single moment in time, the performance was surprisingly similar whether the receiver knew the symbols or not. However, the advantage of knowing the symbols became dramatic when the system observed the targets over a longer period, specifically across multiple time slots. In these multi-slot observations, knowing the symbols allowed the receiver to combine the echoes coherently, effectively stacking the signal to make it much stronger against the noise. Without this knowledge, the observations remained unstructured, and the system could not gain the same benefits from watching longer. The simulations showed that with four time slots of observation, the system with known symbols achieved significantly better radar performance than the system with unknown symbols, while both approaches remained close to the performance of a radar-only system when the communication requirements were moderate.

Ultimately, the study demonstrates that while integrated systems can function effectively even without full knowledge of the transmitted data, the ability to share that information unlocks a powerful capability for long-duration sensing. The proposed designs allow a single network to serve multiple users and detect multiple targets simultaneously, with the radar performance staying remarkably close to what a dedicated radar system could achieve, provided the communication needs are not excessively demanding. This work provides a practical roadmap for building future networks that are not only faster but also inherently aware of their physical environment.