Work for a Member company and need a Member Portal account? Register here with your company email address.
Instead of relying on GPS or cameras, which fail indoors and in the dark, MiFly relies on new localization technology that leverages wireless signals. It can accurately track its motion as it flies throughout an environment.
MiFly leverages millimeter waves (mmWave), which are common in 5G systems and autonomous vehicles. Architecturally, it uses two lightweight mmWave radars mounted on a drone and a custom-designed backscatter anchor deployed in the environment . MiFly achieves precise drone localization using just one mmWave anchor, reducing infrastructure setup and cost.
Dual-Polarization, Dual-Modulation
We co-design a novel dual-polarization and dual-modulation architecture in both our anchor hardware and localization algorithm to enable single-shot 3D localization, as shown in the figure below.
MiFly leverages two radars transmitting different polarizations. Similar to polarized sunglasses, leveraging different polarizations allows us to separate horizontal and vertical signals, enabling full 3D localization. The vertically polarized component of the anchor backscatters the vertically polarized signal, while the horizontally polarized portion backscatters the horizontally polarized radar signal.
Moreover, the two orthogonally polarized components are designed to operate with different modulation frequencies, which enables the radars to isolate their reflections in the frequency domain. Not only does this minimize interference, but it also enables dealing with tilts that naturally occur during flight, which may otherwise cause shifts in polarization.
6DoF Pose Estimation:
Anchor measurements provide 3D estimates of the anchor relative to the radars, these estimates cannot be directly converted to the drone’s position due to ambiguity caused by the drone’s 6DoF (degrees of freedom) motion during flight.
By fusing mmWave radar data with the drone’s Inertial Measurement Unit (IMU), MiFly provides complete six degrees-of-freedom (6DoF) localization.
We implement MiFly on a DJI Mavic 3 Classic drone and tested in across various indoor environments, collecting over 6,600 localization estimates.
Median 3D error: Achieves median errors of 4.8 cm, 1.0 cm, and 3.0 cm in the x, y, and z dimensions, respectively.
Performance in NLOS scenarios: Performs almost equally well in line-of-sight (LOS) and non-line-of-sight (NLOS) scenarios, with a marginal drop in NLOS (∼1 cm)
Low-light & GPS-denied operation: Functions without relying on visual features or GPS, allowing for operation indoors, in the dark, and low-visibility settings.