The Hera framework for fault-tolerant sensor fusion on an Internet of Things network with application to inertial navigation and tracking

Classical sensor fusion approaches require to work directly with the hardware and involve a lot of low-level programming that is not suited for reliable and user-friendly sensor fusion for Internet of Things (IoT) applications. In this master thesis, we have developed Hera, a Kalman filter-based sensor fusion framework for Erlang which offers a high-level approach for dynamic, soft real-time, and fault-tolerant sensor fusion directly at the edge of an IoT network. We use the GRiSP-Base board, a low-cost platform specially designed for Erlang and to ease development by avoiding soldering or dropping down to C. We emphasise on the importance of performing all the computations directly at the sensor-equipped devices themselves, completely removing the cloud necessity. With Hera, the implementation effort is significantly reduced which makes it an excellent candidate for IoT prototyping and education in the field of sensor fusion. We explain the basis of inertial navigation and tracking, and show that with the use of Erlang, GRiSP, and Hera, it is possible to perform simple sensor fusion for position and orientation tracking at a high-level of abstraction. From a fault-tolerance analysis based on fault-injection, we show that Hera gives the strong guarantee to do sensor fusion as long as one GRiSP board is alive. In a first phase, we experiment with Hera to show how we can build a sensor fusion model for a position tracking application and illustrate the benefits of sensor fusion as we add more sensors or increase the model complexity. In a second phase, we attain the limit of the current system by tackling a more challenging 6 degrees of freedom (DOF) inertial measurement unit (IMU). We explain the theory and show surprisingly good results for the attitude and heading reference system (AHRS). Finally, we give information about planned performance improvements, to address certain limitations, as well as possibilities for future work.