Integration of UWB Radar on Mobile Robots for Continuous Obstacle and Environment Mapping

This paper presents an infrastructure-free approach for obstacle detection and environmental mapping using ultra-wideband (UWB) radar mounted on a mobile robotic platform. Traditional sensing modalities such as visual cameras and Light Detection and Ranging (LiDAR) fail in environments with poor visibility due to darkness, smoke, or reflective surfaces. In these vision-impaired conditions, UWB radar offers a promising alternative. To this end, this work explores the suitability of robot-mounted UWB radar for environmental mapping in anchor-free, unknown scenarios. The study investigates how different materials (metal, concrete and plywood) and UWB radio channels (5 and 9) influence the Channel Impulse Response (CIR). Furthermore, a processing pipeline is proposed to achieve reliable mapping of detected obstacles, consisting of 3 steps: 1) target identification (based on CIR peak detection); 2) filtering (based on peak properties, signal-to-noise score, and phase-difference of arrival); and 3) clustering (based on distance estimation and angle-of-arrival estimation). The proposed approach successfully reduces noise and multipath effects, achieving high obstacle detection performance across a range of materials. Even in challenging low-reflectivity scenarios such as concrete, the method achieves a precision of 73.42% and a recall of 83.38% on channel 9. This work offers a foundation for further development of UWB-based localisation and mapping (SLAM) systems that do not rely on visual features and, unlike conventional UWB localisation systems, do not require fixed anchor nodes for triangulation.

💡 Research Summary

The paper presents an infrastructure‑free obstacle detection and environment‑mapping solution for mobile robots that relies on ultra‑wideband (UWB) radar instead of conventional vision or LiDAR sensors. The authors motivate the work by pointing out the limitations of cameras (require good illumination) and LiDAR (susceptible to smoke, dust, reflective surfaces, and transparent objects) in low‑visibility environments such as darkness, fire, or industrial settings. UWB, with its large fractional bandwidth (>20 %) and sub‑nanosecond pulses, offers high range resolution and the ability to operate through obscurants, but most existing UWB‑based localisation systems depend on fixed anchor nodes, which is impractical for unknown or rapidly changing environments.

Hardware and experimental setup
The study uses an off‑the‑shelf Qorvo QM33120WDK1 development kit (IEEE 802.15.4‑compliant) mounted on a TurtleBot 4 platform. The kit provides an omnidirectional transmit antenna and two directional receive antennas on the same chip, enabling the extraction of both the channel impulse response (CIR) and the phase‑difference of arrival (PDoA) between the two receive paths. Data are captured every 0.42 ms, yielding a high‑rate stream of CIR snapshots. Experiments are conducted in a large industrial testbed with three representative obstacle materials—metal, concrete, and plywood—using IEEE 802.15.4 channels 5 (≈3.5 GHz) and 9 (≈4.3 GHz).

Signal‑processing pipeline
The proposed pipeline consists of three sequential stages:

1. Target identification – Peaks in the CIR are detected using configurable width and prominence thresholds. Each peak corresponds to a reflected path; the first peak after the line‑of‑sight (LoS) component typically represents the obstacle of interest.

2. Filtering – For each detected peak, three attributes are computed: (a) peak shape (width, height, location), (b) signal‑to‑noise ratio (SNR), and (c) PDoA derived from the phase difference between the two receive antennas. These attributes are combined into a reliability score; peaks below a score threshold are discarded as multipath‑induced ghosts. This step dramatically reduces false detections while keeping computational load low enough for real‑time operation.

3. Clustering and mapping – Distance is estimated from the peak’s time delay (time‑of‑flight), and azimuth is estimated from the PDoA. The resulting polar coordinates are transformed into Cartesian points. A density‑based clustering (DBSCAN‑like) groups points that belong to the same physical obstacle across successive measurements, producing stable obstacle representations (position and approximate width). The clusters are then rendered into a 2‑D occupancy map that updates continuously as the robot moves.

Results
On channel 9, the method achieves a mean absolute distance error of ≤ 16.7 cm across all materials. For the most challenging low‑reflectivity material (concrete), precision reaches 73.42 % and recall 83.38 %; metal and plywood yield higher accuracies (> 90 %). These figures surpass prior works that rely on accumulation‑based CIR (ACIR), singular‑value decomposition, or deep‑learning classifiers, which typically report 55‑70 % detection accuracy in similar settings. The authors also release the full dataset, processing code, and a demonstration video, facilitating reproducibility and further research.

Discussion and limitations
The approach is limited to a detection range of roughly 10 m and provides only 2‑D azimuth estimates; the angular resolution is constrained by the small antenna spacing on the chip (tens of degrees). Complex indoor geometries (curved walls, glass) still generate residual multipath that can cause occasional ghost clusters. The authors suggest future work on multi‑antenna arrays for finer AoA resolution, sensor fusion with IMU/odometry for full 3‑D SLAM, and adaptive machine‑learning models to automatically tune peak‑detection thresholds in varying environments.

Conclusion
By combining low‑cost, standards‑compliant UWB hardware with a lightweight, three‑stage signal‑processing chain, the paper demonstrates that mobile robots can reliably detect and map obstacles in environments where cameras and LiDAR fail. The solution is lightweight, power‑efficient, and does not require any pre‑deployed infrastructure, making it attractive for search‑and‑rescue, industrial inspection, and autonomous logistics where rapid deployment and robustness to visual obscurants are essential.