Skip to main content
Back

Two Drones Map Each Other's Wi-Fi Blind Spots in One Flight

A Czech-Moroccan robotics team shows how two drones can learn each other's antenna radiation patterns mid-flight, hitting 4.56 dB RMS prediction error.

Paper, Full Access
  • New method learns antenna radiation patterns of two quadrotors simultaneously from a single joint calibration flight.
  • Achieves 4.56 dB RMS prediction error using spherical harmonics with roughly 200 parameters per antenna.
  • Beats a 400-nearest-neighbor baseline (4.82 dB) while also decoupling the joint gain into per-UAV patterns.
  • Calibration trajectory uses vertical circles with stepwise heading changes to cover azimuth-elevation space.
  • With only two UAVs the isotropic gain is unidentifiable; a third aircraft removes the rank deficiency.
  • Full paper: arxiv.org/abs/2604.02827.

Researchers at the Czech Technical University in Prague and Mohammed VI Polytechnic University have developed a field-calibration method that estimates the Wi-Fi radiation patterns of two quadrotors during one coordinated flight. The calibration method separates the directional gain of aircraft carrying different payloads, allowing communication-aware planners to account for heading-dependent signal loss without an anechoic chamber.

The airframe rewrites the antenna

Communication-aware path planning, swarm control, and airborne relay systems rely on models of received signal power. Multirotor links can appear erratic because the frame, battery, wiring, and payload alter the antenna’s electromagnetic field. Each aircraft therefore behaves as a complex antenna whose gain changes with azimuth and elevation.

The free-space Friis equation estimates received power from transmit power, propagation loss, and the directional gains of the transmitting and receiving antennas. Distance and transmit power are usually known. The two antenna-gain terms are harder to obtain because they change whenever the aircraft configuration changes.

On the team’s platform, aircraft heading produced repeatable signal-strength differences approaching 10 dB. The paper also cites prior work reporting losses of up to 20 dB. A 10 dB reduction leaves one-tenth of the received power, enough to consume a healthy link margin and cause a dropout.

Anechoic-chamber measurements can characterize a fixed configuration, but a new camera, sensor pod, radio mount, or battery position can invalidate the result. Field calibration provides a faster way to update the model after hardware changes.

Two Holybro X500 V2 quadrotors performing the coordinated calibration flight
The coordinated trajectory samples the relative azimuth and elevation of both aircraft.

Two drones, one calibration flight