Embedded software design of internet of things positioning device

Abstract

KAILOS system (\url{https://kailos.io/}) aimed to provide an integrated indoor-outdoor navigation experience. However, KAILOS suffers badly from the inherent problems of infrastructure-dependent methods and the restriction on smartphone applications. KAILOS tag - a dedicated Internet-of-Things (IoT) positioning device - was created to overcome those challenges. There were three important requirements that needed to be fulfilled when designing its embedded software: algorithm performance, real-time capability, and power efficiency. We proposed an efficient operation cycle and an effective yet accurate Pedestrian Dead Reckoning (PDR) algorithm to address these requirements. The operation cycle relied on priority-based scheduling and sensor time synchronization methods. The PDR algorithm involved a dynamic gyroscope bias update technique that automatically detects the stationary state of the sensor using acceleration differential to calculate the bias of the gyroscope in a real-time manner. Additionally, to maintain the consistency of the bias values at the beginning of every movement, we gathered the sensing data and calculated the bias using a scalable window. Final results showed a low packet loss rate of 3.3\% for Inertial Measurement Unit (IMU) sensor sampling at 100 Hz, and the median of server latency reached only 2.1 seconds maximum over a 30-minute duration. The current consumption of the tag when in fully functioning mode was 175 mA on average. For the PDR performance, the detected step error remained low at 1.5\% while the traveled distance accuracy achieved a minimum of 98.43\%. The end-to-end error was also below 1.6\%, and the estimated trajectories showed significant improvements over the traditional PDR approach. The proposed solutions could be applied to any other dedicated IoT platform that wants to introduce both multiple sensor integration and real-time ability to their system.