We shows how changing antenna layouts together with transmitted signals can improve target localization in networks that share communication and sensing functions. Simulations at 28 GHz found a 5.51 dB gain in a theoretical localization error bound over a compact fixed-array baseline. The analysis also identifies missing angle information in single-pattern probing and shows how calibration errors can erode the benefit.
Wireless networks can use reflected radio signals to locate objects while also carrying data. Allowing antennas to move adds another way to improve these sensing functions, but the benefit depends on what information the echoes actually contain. A study by Changhao He at King Abdullah University of Science and Technology (KAUST) examines this question and develops a method for jointly choosing antenna layouts and transmitted signals.
The work considers cell-free integrated sensing and communication, in which distributed access points cooperate to serve users and observe a target from different directions. Each access point can adjust the positions of its antenna elements. These adjustments change how the network illuminates a target and receives its echoes, creating opportunities to improve localization while meeting communication quality and power requirements.
The analysis begins with a source of uncertainty that can make sensing predictions too optimistic. The strength and phase of a reflected signal are generally unknown. When these unknown quantities are included in the model, a probing signal with only one independent spatial pattern provides no transmit-angle information for the modeled link. At least two independent patterns are necessary, although that alone does not guarantee useful angle information. The finding helps distinguish apparent sensing information from information that can actually identify a target.
Using this corrected information model, the study alternates between optimizing communication and sensing signals and adjusting the transmitting and receiving array shapes. It evaluates the designs with a Cramér–Rao bound, a theoretical lower bound on localization error variance. In 48 paired simulations at 28 GHz, the joint design reduced the mean position error bound from 27.57 to 7.77 square metres relative to signal optimization with a compact fixed layout. The average gain computed within paired trials was 5.51 dB. This comparison combines the benefit of occupying more of the allowed aperture with that of choosing its shape.
Further comparisons tested whether the improvement depended on a favorable choice of baseline. The joint design outperformed 20 prespecified random layouts across all 960 paired layout-and-trial comparisons. Separate maximum-likelihood localization simulations checked whether the theoretical ranking translated into estimation performance. At the tested higher-signal checkpoints, the joint design’s root-mean-square position error stayed within 0.45 dB of its corresponding bound.
Calibration sets a practical limit on these gains. Uncertainty in array orientation substantially worsened localization in the evaluated setting, and redistributing the signal covariances offered little improvement. The results point to a design priority: accurate geometric calibration is needed to realize the benefit of antenna reconfiguration. The evidence comes from mathematical analysis and numerical simulations; hardware validation remains a next step.
This paper “Nuisance-aware geometry-reconfigurable cell-free ISAC: identifiability, joint design, and calibration limits” was published in Advanced Information and Communication .
He C. Nuisance-aware geometry-reconfigurable cell-free ISAC: identifiability, joint design, and calibration limits. Adv. Inf. Commun . 2026(4):0014, https://doi.org/10.55092/aic20260014 .
Advanced Information and Communication
Computational simulation/modeling
Not applicable
Nuisance-aware geometry-reconfigurable cell-free ISAC: identifiability, joint design, and calibration limits
29-Oct-2026