One of the most widely used techniques for locating a mobile object such as a drone inside a building without satellite signals is to place ultra wideband (UWB) beacons on the ground. The radio-equipped flying machine then uses these beacons as reference points to determine its position.
“However, in a warehouse for example, radio signals can be disrupted by obstacles such as shelves," explains Christophe Villien, signal processing researcher at CEA-Leti. “What's more, location accuracy decreases as the drone moves further away from the area covered by the beacons: this is known as the geometric dilution of precision." And since drones rise up into the air, they will always tend to deviate from these markers, which are located on the ground.
From UWB radio to radar
To overcome these limitations, Christophe Villien has been working on a new indoor localization method: “This work was inspired by real-life use cases from an industrial company that provides drones for warehouse inventories or the 3D inspection of aircraft hulls."
His idea was to improve the accuracy of the drone's localization by exploiting data provided by the equipment already in place without having to change the hardware. To achieve this, Christophe Villien hijacked the primary function of the device's UWB radio antennas:
“They are designed to provide a single measurement: the distance between them and the beacons on the ground. But in reality, they have access to a much wider range of data. Usually, this data set is processed with the goal of determining the distance to a beacon, but it can also be used to measure a signal's echo from surrounding objects."
This analysis then enables them to evaluate the distance to these obstacles, much like radar.
Combining radar and UWB beacons
However, this approach faces a major difficulty because of the multitude of echoes being received.
“Our aim is to locate the drone in relation to known, fixed obstacles. So we have to sort out the echoes and retain only those of interest to us," says Christophe Villien. “To do this, we couple our radar technique with the distances measured to ground beacons in order to provide a composite estimate of the device's position."
The combination of these two methods has significantly improved the accuracy and robustness of localization, with for example, an average vertical error of 40 cm in field tests (compared to 96.4 cm with only UWB beacons).
These results attracted the attention of the IPIN (Indoor Positioning & Indoor Navigation) community. Christophe Villien presented his work at the fourteenth edition of this international conference, held in Hong Kong (China) in October 2024, where he was awarded fourth prize for the "Best Paper Awards".
While the method developed is particularly well suited to the industrial applications that were initially targeted, Christophe Villien is considering new horizons:
“We have proposed a thesis based on the same building blocks, but with the objective of dispensing with ground-based UWB beacons altogether. Thanks to the introduction of AI, we would be able to use only echo analysis for localization and tracking. This could open the way for use in the field, for example to provide soldiers with information within a building."