Fast Real-World Implementation of a Direction of Arrival Method for Constrained Embedded IoT Devices

Abstract

Direction of arrival (DOA) methods are found in many applications, and in the case of the Internet of Things (IoT), it is used for indoor localization. However, the implementation of DOA in IoT devices poses a real challenge, since they are computationally expensive complex numerical methods that could easily lead to resource starvation, unacceptable execution time, and rapid depletion of batteries of small constrained embedded systems typically found in IoT networks. This paper contributes to alleviating that problem, it presents a fast low-power optimized version of a DOA method called Unitary TLS ESPRIT. The optimization exploits the radio communication system design to avoid two time-consuming executions of eigendecomposition, and instead, it applies two simple Power Method algorithms. The result is a lightweight version of ESPRIT that can attain sub-millisecond execution time. To prove the solution’s viability, we carried out experiments on energy consumption, memory footprint, accuracy, and execution time for three floating-point formats in a commercial constrained embedded IoT device series without any operating system and software layers. Experiments show the solution satisfies the hardware requirements and the floating-point precision fully operated by the Floating-Point Unit is found to be the best option.