Where We’re Going, We Don’t Need GPS
MIT's MiFly uses mmWave signals and a small tag to help drones navigate precisely in GPS-denied locations, like tunnels and urban canyons.
To get their bearings and make autonomous navigation possible, modern drones most commonly rely on GPS signals. And no wonder, because GPS is a very practical way to quickly determine the location of a drone to within a few millimeters of its true position. As long as GPS signals are available, that is. In indoor environments, or in urban settings where buildings and other obstructions block GPS, these signals cannot be counted on for localization. This means that drones working in warehouses, or delivering packages in cities, for instance, need alternative methods for determining their position.
One potential method, called MiFly, was recently proposed by a team led by researchers at MIT. Unlike existing alternatives that rely on computer vision or LiDAR — which can struggle in low-light or featureless settings — MiFly employs millimeter-wave (mmWave) radio signals to enable precise self-localization with minimal infrastructure. It was shown that this system is capable of precisely locating drones in GPS-denied environments such as tunnels and urban canyons.
MiFly is built around a simple yet effective concept — it enables a drone to locate itself by using radio waves reflected from a single small tag placed in its environment. This tag, which requires very little power, can be affixed to a wall like a sticker, making the system cost-effective and easy to deploy.
MiFly employs a dual-polarization, dual-modulation system. The drone is equipped with two radars — one mounted horizontally and one vertically — which emit mmWave signals. The tag reflects these signals back to the drone, but with a twist: it modulates the signals in a way that differentiates them from other environmental reflections. This allows the drone to isolate and interpret the signals reflected by the tag, which can be leveraged in calculating its precise location in three dimensions.
To achieve even greater accuracy, MiFly integrates data from the drone’s inertial measurement unit (IMU), which tracks changes in acceleration, rotation, and orientation. By fusing radar-based location estimates with IMU data, MiFly enables the drone to determine its exact position and movement with six degrees of freedom — meaning it can account for roll, pitch, yaw, and lateral movement.
The researchers tested MiFly in a variety of challenging environments, including dark rooms, warehouse-like spaces, and underground tunnels. In these trials, the system was able to localize the drone to within just 7 centimeters of its true position in many cases — a level of precision comparable to state-of-the-art GPS-based systems.
Even in scenarios where the tag was visually blocked from the drone’s line of sight, MiFly maintained reliable positioning up to six meters away. The researchers believe that this range could be extended further with improvements in radar and antenna design or by using more powerful signal amplifiers.
At present, the team is working to integrate MiFly into a full-fledged autonomous navigation system. By combining MiFly with advanced flight planning algorithms, they hope to enable drones to not only determine their position but also plan and execute flight paths independently.
