Navigation of unmanned vehicles using Bluetooth

This page describes multiple projects related to navigation of autonomous vehicles using Bluetooth, in a project called PARNAV.Collaboration between multiple students on related projects is highly encouraged. See also this link for other projects within PARNAV offered by Assoc. Prof Torleiv H. Bryne.

Background

Bluetooth is a well-known wireless technology standard that enables communication between many of our smart devices. However, recent and upcoming additions to the Bluetooth standard enable new use cases; navigation, using range and/or directional information. In version 5.1 of the Bluetooth standard, the constant tone exchange (CTE, a constant sine wave) was added to the end of the Bluetooth package. By measuring the phase of this sine wave using multiple antennas in different positions, the direction to the signal source can be estimated.

The range between the transmitter and the receiver can be found in at least three different ways:

1. Using signal strength (RSSI) is typically the least accurate,
2. Measuring the time it took for the data to travel (time of flight) improves this somewhat, to an accuracy of 2-4 meters.
3. The most accurate method, multi-carrier phase differencing, considers the phase, which also is a function of the distance between the sender and transmitter, enabling accuracies down to 30-50 cm. This was introduced in Bluetooth version 6.0, from August 2024.

While the focus of the industry and the literature seem to be on indoor use of Bluetooth, the underlying technology also has merit in outdoor navigation of unmanned vehicles, like landing of drones and docking of autonomous ships, see the below videos:

Net landing	Ship docking

Both videos represent scenarios where it is critical to know the exact position of the vehicle, so relying on a single position sensor, like GPS, makes the system vulnerable and error-prone.

Scope

There are many possible areas of focus within this project, that can be adapted to the interests and competence of the candidate. This includes, but is not limited to

• AI for Signal Processing: Can the multipath/interference experienced in e.g. the landing scenario be predicted/learned and mitigated? Preliminary results indicate “yes”. Some ideas:
  • use of machine learning or classical methods to identify and mitigate multipath effects. The preliminary results, fitting Fourier series, gives promising results, but can we get an equally good fit with less data?
  • how can we fly to best observe and learn the multipath effects in an area? Both when we have GNSS and when we don’t.
• Sensor Fusion & Autonomous Systems: How can Bluetooth range and direction measurements be combined with other measurements to estimate the position, velocity and attitude of autonomous vehicles? One option could be to use a Kalman filter to fuse the information from inertial sensors (accelerometer and gyroscope) with multiple range measurements and other sensors common in drone payloads (barometer/altimeter, GPS). Some ideas:
  • improve measurements by using velocity feedback from the KF to compensate for Doppler effects?
  • use feedback from previous estimates to initiate the search for the new measurement. How to ensure stability?
  • use of neural networks to detect anomalies in the power spectrum
  • comparison of various direction finding algorithms on real Bluetooth data from a drone
  • Phase-based ranging and direction finding share the same physical radio. What is the optimal balance between range and direction measurements?
  • Accuracy of phase-based ranging in a stationary scenario improves when using more frequencies. But measuring more frequencies takes more time, so the robot has moved (slightly) in the meantime, leading to lower accuracy for agile robots. For a given robot dynamics, what is the optimal trade-off?
• Search and Rescue with Multi-Agent System: In a search and rescue situation, two drones equiped with Bluetooth arrays are trying to locate a missing person through the Bluetooth signal from his phone. What is the best way to do this? How should the drones move?
  • consider both drones as one big array (to increase the aperture and thus the accuracy)?
  • range and direction information from both drones, centrally processed?
  • de-centralized processing?
• Smart Warehouse Robotics: AutoStore, a world-leading company in warehouse robots, is a partner in the project. They are interested in finding the position of their robots, particularily those that have broken down/derailed. Some related challenges:
  • how to exploit information about the grid/map to improve navigation accuracy
  • what is the optimal placement of arrays/beacons along the grid to maximize navigation accuracy and minimize cost? Can we save cost/equipment by positioning idle robots to improve accuracy, or actively position robots to find the one that has derailed?
  • relative motion measurement (we moved 1mm) are very accurate in BT ranging, but absolute suffers from bias. Can this be mitigated with a KF?

Please indicate what aspects of the project that caught your attention.

Regardless of the focus, experimental validation will be important, where flight testing using one of our drones is a natural step on the way.

Proposed tasks

The tasks will vary based on the focus of the project, as decided by the student (you) and the supervisor (me), but the below steps are common:

• Perform a literature study on the state of the art within radio-based navigation
• Familiarize with the capabilities and limitations of relevant, existing development kits and software libraries for Bluetooth range and/or direction measurements
• Discuss the results with a critical eye, and conclude the work in a written report

What you will learn

This project offers hands-on experience with cutting-edge wireless positioning technology and autonomous systems. Depending on your chosen focus and background, you will develop skills in:

• State-of-the-art wireless technology: Work with Bluetooth 6.0 (released August 2024) and centimeter-level positioning before it becomes mainstream
• Signal processing & RF engineering: Antenna arrays, phase-based measurements, direction finding algorithms, and multipath mitigation
• Sensor fusion & state estimation: Kalman filtering, multi-sensor integration, and navigation algorithms for autonomous vehicles
• Machine learning for signal processing: Apply ML or classical methods to improve measurement accuracy in challenging RF environments
• Autonomous systems in practice: Emphasis on real-world testing with hands-on hardware validation, e.g. flight tests with the NTNU UAVlab.
• Industry collaboration: Work with tech from Norwegian, world-leading companies (Nordic Semiconductor, AutoStore) with potential career opportunities

No prior expertise in all areas is expected—your interests will help shape the project focus.

Contact

Contact supervisor Kristoffer Gryte. Other people involved in the project, depending on the chosen focus, are

• Torleiv H. Bryne (Assoc. Prof. NTNU ITK)
• Kimmo Kansanen (Prof. NTNU IES)
• Carsten Wulff (Assoc. Prof. NTNU IES/Wireless Group Manager Nordic Semiconductor)

References