Smarter Anchor Placement Boosts Outdoor UWB Tracking Accuracy

Ultra-wideband radio has long promised centimeter-level positioning, but the technology’s real-world performance depends on something deceptively mundane: where you put the antennas. A new study published in Complex & Intelligent Systems by Byeonghyun Lee, Jeik Choi, JungHun Choi, and Ki-Yong Oh of Hanyang University and Hyundai Motor Company’s ZERO1NE Incubation Team tackles this problem head-on, presenting a framework that optimizes the placement of UWB anchor stations for tracking a pedestrian surrogate in outdoor environments. The work, published as an open-access article on 13 October 2026, arrives at a moment when autonomous vehicles, delivery robots, and mobile test platforms increasingly need to know exactly where people are around them, even when satellite navigation signals are unreliable.

The core challenge the researchers address is that UWB positioning accuracy is not a fixed property of the hardware. It varies dramatically with the geometry of the anchor network, the height at which anchors are mounted, the orientation of their antennas, and the propagation conditions of the surrounding environment. In outdoor settings, signals can bounce off buildings, be blocked by obstacles, or degrade over long distances, creating non-line-of-sight conditions that corrupt the time-of-flight measurements on which UWB ranging depends. A poorly placed anchor network can turn a theoretically precise system into one with errors measured in meters rather than centimeters, and until now there has been no systematic, experimentally validated method for choosing anchor positions outdoors.

The framework proposed by the team rests on three pillars. The first is a rigorous quantification of localization error that allows the researchers to estimate three-dimensional positioning errors for any candidate anchor configuration under realistic outdoor propagation conditions. Rather than relying on idealized geometric metrics alone, the framework incorporates the distortions that actually afflict ranging measurements in the field, such as the biases introduced when signals travel indirect paths or weaken with distance. This means the optimization process evaluates configurations not as abstract geometric arrangements but as physical systems operating in messy, real-world radio environments.

The second pillar quantifies radio-frequency coverage. The researchers apply min-max normalization to received-signal-strength measurements, producing a coverage metric that reflects how reliably each point in the operating area can communicate with the anchor network. Crucially, this quantification incorporates the effects of antenna orientation and anchor height, the two deployment variables that field engineers most often have to guess at. By folding these factors into the coverage model, the framework can evaluate how stable the communication link will be across diverse installation scenarios, from anchors mounted low on tripods to units fixed high on poles or structures.

The third pillar is Bayesian optimization, a machine-learning technique that efficiently searches complex, expensive-to-evaluate design spaces by building a probabilistic model of the objective function and choosing each next candidate based on the trade-off between exploring unknown regions and exploiting known good ones. Here, the researchers construct a hybrid objective function that simultaneously accounts for coverage robustness and localization accuracy, allowing the optimizer to find anchor configurations that balance the two goals rather than sacrificing one for the other. The weight parameter governing this balance was itself tuned through a sensitivity analysis conducted in a Gazebo-based virtual environment, where six weight values ranging from zero to one were tested. The analysis revealed a highly nonlinear relationship between the weight and positioning error, with a value of 0.01 achieving the lowest root mean square deviation of 1.28 meters in the virtual trials, and that value was adopted throughout the main optimization framework.

To validate the approach, the team ran controlled experiments examining ranging distortions and signal degradation, verifying that each component of the framework contributes meaningfully to pedestrian-surrogate localization performance. They then carried out extensive field experiments across diverse outdoor environments, comparing the optimized anchor configurations against alternatives. The results confirmed both communication stability and improved localization accuracy under the configurations produced by the proposed framework, demonstrating that the method translates from simulation and controlled tests into genuine field performance.

One of the most striking findings comes from the paper’s appendix, where the authors report an ablation study designed to isolate the contribution of the anchor placement itself from the post-processing algorithms applied afterward. The full localization pipeline uses a moving average for smoothing and a velocity constraint to prevent excessive positional drift in geometrically poor configurations. The study compared three placement strategies, one based on position dilution of precision, one on the condition number of the geometry matrix, and one on the proposed method, across four algorithm variations from a standalone least-squares solver to the full pipeline. The proposed placement consistently achieved the lowest error in every case: 2.46 meters with the standalone least-squares approach, compared with 3.71 meters for the condition-number case and 4.10 meters for the PDOP-based case. With the full pipeline, the proposed method recorded 1.54 meters against 2.26 and 1.69 meters for the alternatives. The consistent gap confirms that the accuracy gains originate fundamentally from the optimized anchor geometry, with post-processing merely stabilizing extreme outliers.

The technical vocabulary of the study reflects the sophistication of modern UWB positioning. The system relies on time-of-flight ranging, using a carrier frequency offset-corrected single-sided two-way ranging method to estimate distances between the moving tag and fixed anchors. Channel impulse response data and received signal strength feed into the coverage and error models, while dilution-of-precision metrics provide the classical geometric baseline against which the new method is judged. By combining these established measurement techniques with Bayesian optimization, the framework bridges the gap between radio-frequency engineering and machine learning, treating anchor placement as a stochastic optimization problem rather than a rule-of-thumb engineering decision.

The practical implications extend well beyond the laboratory. The research was supported by Hyundai Motor Company’s ZERO1NE Incubation Team through the WhereB Project, along with funding from Korea’s Ministry of the Interior and Safety and the Seoul ANCHOR program, suggesting automotive applications in which vehicles or test platforms must track pedestrians with high reliability. A pedestrian surrogate, in this context, is a stand-in for a real person, allowing safety systems to be tested against realistic human movement without exposing anyone to risk. Accurate outdoor localization of such surrogates is essential for validating collision-avoidance systems, robot navigation, and infrastructure-assisted safety features, particularly in environments where global navigation satellite systems struggle, such as urban canyons, parking structures, and areas with heavy foliage.

Perhaps the most valuable contribution of the work is its generality. The authors emphasize that the framework provides novel guidelines for outdoor UWB anchor placement that can adapt across diverse deployment environments and operational conditions, rather than producing a single fixed solution. Because the objective function weighs both coverage robustness and localization accuracy, and because the underlying error and coverage models account for antenna orientation and mounting height, the same optimization machinery can be redeployed in a new site with new measurements. The authors note that future work will pursue dynamic weight optimization adapted to specific deployment environments, hinting at systems that could retune themselves as conditions change. For a field that has often depended on intuition and trial-and-error, the study offers something rarer: a principled, experimentally grounded recipe for getting ultra-wideband positioning right the first time, with errors low enough to matter for the safety-critical systems that increasingly share our streets.

Subject of Research: Optimization of ultra-wideband anchor placement for outdoor pedestrian localization

Article Title: Anchor placement optimization in outdoor UWB systems for pedestrian-surrogate localization

Article References: Lee, B., Choi, J., Choi, J., & Oh, K.-Y. (2026). Anchor placement optimization in outdoor UWB systems for pedestrian-surrogate localization. Complex & Intelligent Systems. https://doi.org/10.1007/s40747-026-02542-7

Image Credits: AI Generated

DOI: 10.1007/s40747-026-02542-7

Keywords: UWB, anchor placement, Bayesian optimization, localization, RF coverage, RSS, pedestrian surrogate, outdoor positioning, time-of-flight ranging, dilution of precision, autonomous vehicles, wireless sensing