Combining GNSS, Bluetooth® Angle of Arrival, and inertial sensing could provide the relative positioning needed to recover autonomous drones on moving vessels at sea.
Fixed-wing drones are well suited to maritime search and rescue. They can travel long distances, cover large areas quickly, and give rescue crews an aerial view far beyond the horizon. But there is a practical problem: once the mission is over, how does the drone return to a small rescue vessel that is itself moving with the waves?
On land, autonomous recovery can rely on a stationary landing location. At sea, the target is continuously translating and rotating. Pitch, roll, and yaw alter the orientation of the vessel, while wind and waves affect both the aircraft and its recovery platform.
Research carried out at Lund University as part of the Vinnova-funded NextG2Com competence center explored whether combining several complementary positioning technologies could provide an answer.
Working with u-blox and the Swedish Sea Rescue Society (SSRS), master’s student Beau Forrez developed a ship-mounted radio guidance system combining GNSS, Bluetooth® Direction Finding, and inertial measurements.
The result demonstrates an important principle for autonomous systems: sometimes the best positioning solution is not one technology, but several technologies working together.
For search-and-rescue organizations such as SSRS, fixed-wing unmanned aerial vehicles (UAVs) offer an attractive way to extend surveillance beyond the range of a rescue vessel.
Compared with many multirotor drones, fixed-wing aircraft can remain airborne longer and efficiently search large areas. This makes them particularly useful when every minute matters.
But their endurance advantage creates another challenge. Recovering a fixed-wing aircraft generally requires more space and more precise guidance than recovering a hovering multirotor.
Returning the aircraft to shore reduces the useful portion of its flight time. Recovering it directly aboard the rescue vessel could therefore significantly extend the area and duration of maritime search operations.
Fixed-wing UAV used to explore autonomous recovery concepts for maritime search-and-rescue operations.
Doing that autonomously requires the drone to know not simply where it is, but where it is relative to a recovery point that is continuously moving and rotating.
That distinction is important.
GNSS provides an excellent global position reference. But during the final approach, a drone also benefits from precise information about the direction of the vessel relative to the aircraft. Bluetooth Direction Finding can provide this local angular information.
The challenge is bringing those measurements into the same frame of reference.
The architecture investigated in the thesis uses complementary positioning technologies for different parts of the recovery operation.
GNSS provides absolute positioning and supports guidance while the drone is farther away from the vessel. As the aircraft approaches, Bluetooth 5.1 Angle of Arrival (AoA) measurements provide relative directional information that can support terminal homing.
GNSS provides long-range navigation, while Bluetooth Angle of Arrival (AoA) provides directional guidance during the final approach to the recovery vessel.
In the experimental implementation, the system brought together commercially available u-blox technology, including:
Rather than asking one sensor to solve every positioning problem, the architecture takes advantage of their complementary characteristics.
GNSS establishes the position of the platforms in a global coordinate system. Bluetooth AoA provides information about the direction from which a Bluetooth signal arrives at an antenna array. An inertial measurement unit (IMU), meanwhile, provides the information needed to understand how the vessel itself is moving.
Together, these measurements can build a more complete picture of the drone’s position relative to the vessel.
There is, however, another complication.
Bluetooth Angle of Arrival relies on an antenna array to determine the direction of an incoming radio signal.
In a static application, the orientation of the antenna array can be considered fixed. If an incoming signal moves relative to the array, it is reasonable to infer that the transmitter has moved.
A vessel at sea changes this assumption completely.
Imagine a drone maintaining exactly the same position while the ship beneath its Bluetooth antenna rolls to one side. From the antenna’s perspective, the apparent arrival angle changes even though the drone has not moved.
The same happens with pitch and yaw.
A direction-finding system that interpreted these measurements directly could therefore mistake movement of the ship for movement of the drone.
This is where inertial sensing becomes essential.
The project’s key technical contribution was an IMU-compensated Bluetooth AoA navigation architecture.
Instead of sending the raw Bluetooth angle measurements directly into the navigation filter, the system first accounts for the vessel’s orientation.
IMU measurements describe the rotation of the vessel and antenna array. The Bluetooth AoA measurements can then be transformed from the vessel’s moving body frame into a stable Earth-fixed North-East-Down (NED) coordinate frame.
Conceptually, the processing chain looks like this:
Bluetooth AoA measurement -> vessel attitude compensation -> NED coordinate transformation -> sensor fusion -> relative drone position
This transformation separates two effects that otherwise appear together in the radio measurements: the actual movement of the drone and the apparent movement caused by the vessel rotating underneath the antenna array.
In experiments involving wave-like vessel motion, this approach successfully separated the drone’s true approach direction from disturbances caused by platform movement.
That makes the concept relevant well beyond maritime drone recovery. Robots, vehicles, industrial machinery, and other moving platforms can all encounter the same fundamental problem when direction-finding sensors are mounted on a platform whose orientation changes.
Correcting the reference frame solves only part of the problem.
GNSS, Bluetooth AoA, and IMU sensors do not necessarily produce measurements at the same time or update rate. They also measure different physical quantities and have different uncertainty characteristics.
The navigation system therefore needs to combine asynchronous information into one continuously updated estimate.
For this, the project used an Extended Kalman Filter (EKF).
The EKF maintains an estimate of the relative state of the approaching drone and updates that estimate whenever new sensor measurements become available. GNSS contributes global position information, Bluetooth AoA contributes relative directional observations, and inertial measurements make it possible to compensate for the vessel’s changing attitude.
The result is more useful than simply averaging several sensors together. Sensor fusion allows the system to exploit the strengths of each measurement while accounting for its uncertainty and limitations.
This illustrates why hybrid positioning is becoming increasingly important for autonomous machines.
The question is no longer simply, “Which positioning technology should I use?”
Increasingly, it is, “Which combination of positioning technologies provides the information my autonomous system needs at each stage of operation?”
This leads naturally to a two-stage recovery architecture.
When the UAV is far from the vessel, GNSS is the primary source for navigation. Both platforms can determine their absolute positions, allowing the aircraft to navigate toward the vessel over comparatively long distances.
As the UAV approaches the recovery zone, Bluetooth Direction Finding adds local relative information.
The two technologies therefore solve different but complementary parts of the same problem:
GNSS answers: Where are drones and vessels globally?
Bluetooth AoA answers: From the receiving antenna’s perspective, in which direction is the drone?
The IMU answers: How is the receiving platform oriented right now?
Combining these measurements produces the relative navigation information needed to guide an aircraft toward a moving recovery point.
The system was evaluated using both Software-in-the-Loop simulations and physical human-in-the-loop experiments.
The results demonstrated that transforming the Bluetooth AoA measurements into the NED frame before sensor fusion could effectively compensate for vessel movement.
In physical testing, the system maintained tracking errors of approximately one meter. Simulation produced sub-meter mean errors under evaluated conditions.
Most importantly, the system continued to produce a stable relative navigation estimate while the receiving platform was subjected to motion intended to represent the disturbances experienced aboard a vessel.
Full autonomous aircraft recovery was outside what could ultimately be demonstrated because of limitations in the available drone hardware. The work should therefore not be interpreted as a completed autonomous maritime landing system.
Instead, it validates one of the critical building blocks required to create one: robust relative navigation between an approaching UAV and a moving recovery platform.
Another interesting aspect of the project is the hardware used to build the prototype.
Rather than relying on specialized aerospace positioning equipment, the system was assembled using commercially available GNSS and Bluetooth Direction Finding technology.
Fixed-wing UAV platform used in the experimental evaluation of the ship-mounted recovery guidance system.
This included u-blox XPLR-AOA-3 kits and ANT-B10 antenna arrays for Bluetooth Direction Finding, C209 tags as Bluetooth transmitters, and ZED-F9R GNSS receivers.
That matters because autonomous systems increasingly need positioning architectures that can balance performance with size, power consumption, availability, and cost.
Bluetooth Direction Finding is particularly interesting in this context. Bluetooth is already widely deployed as a low-power wireless technology. Adding directional information creates possibilities for applications where knowing that a device is nearby is not enough, the system also needs to know where it is coming from.
Combining that relative information with GNSS opens another set of possibilities.
Maritime UAV recovery is a demanding demonstration case, but the underlying positioning architecture is much broader.
Autonomous robots may move between environments where satellite visibility changes. Vehicles may need global navigation for most of their journey but precise relative positioning when approaching infrastructure. Mobile machinery may need to locate equipment relative to itself even while the host platform is moving.
Asset tracking systems can similarly benefit from combining global and local positioning rather than forcing one technology to cover every operating condition.
The NextG2Com project demonstrates a general design principle: global positioning and local radio positioning do not have to compete. They can complement each other.
GNSS provides the global reference. Short-range radio technologies provide additional relative observations. Inertial sensors describe platform motion. Sensor fusion connects these measurements into a common navigation solution.
As autonomous systems become more capable, architectures of this kind could become increasingly important.
The thesis also points toward several ways the concept could be extended.
High-precision RTK GNSS could improve the global relative position estimate. Dual-antenna GNSS heading could provide another independent measurement of vessel orientation. More adaptive sensor-fusion algorithms could respond dynamically as measurement quality changes during an approach.
New short-range positioning techniques offer another avenue.
Bluetooth Channel Sounding, introduced with Bluetooth Core 6.0, provides a new mechanism for secure, fine-ranging applications. Future systems could potentially combine ranging with directional measurements, adding another dimension to short-range relative positioning.
Together, these technologies point toward navigation systems that select and combine measurements according to the environment rather than depending on a single positioning source.
For maritime applications, the next major step is clear: integrating the positioning architecture with the aircraft’s flight controller and demonstrating closed-loop autonomous recovery on a moving vessel.
The project emerged from a practical question raised by the Swedish Sea Rescue Society: how can long-range drones spend more time searching and less time returning to shore?
The answer requires more than an aircraft capable of flying long distances. The entire operational cycle has to work, including getting the aircraft safely back.
By combining GNSS, Bluetooth Direction Finding, and inertial sensing, the NextG2Com collaboration has demonstrated a promising technical foundation for solving that final part of the mission.
The work also shows the value of collaboration between academia, technology providers, and end users. Lund University brought positioning and communications research, u-blox provided commercially available GNSS and short-range radio technology, and SSRS provided a demanding real-world use case where improved autonomous navigation could ultimately translate into greater search coverage and faster rescue operations.
The resulting architecture is still a research prototype. But its underlying idea is straightforward: when autonomous systems operate in dynamic environments, positioning technologies become more powerful when they work together.
Read the thesis: Ship-Mounted Radio Guidance System for Fixed-Wing Drone Recovery, Beau Forrez, Lund University, 2026.
Get in touch for orders or any queries: sales@rfdesign.co.za / +27 21 555 8400
Courtesy of u-blox