Evaluation of low-cost inertial measurement units for application in high-volume autonomous vehicles

Abstract:
The autonomous navigation of automobiles, airplanes or robots have gained substantial interest in engineering community. For tasks in outdoor environments, the sensor data typically rely on long-time accurate Global Navigation Satellite System (GNSS) data coupled with a high frequency precise short-term accurate inertial measurement unit (IMU). The major demerits are the performance degradation during signal outages for the GNSS and the sensor’s errors that cause accuracy degradation for the IMU with respect to time. Requirements in terms of accuracy, size, update rate or budget for different applications lead to the need for different sensor combinations. This paper deals with investigation and analyzing the static and dynamic behaviour of a low-cost Micro-Electromechanical System (MEMS) IMU "Lsm9ds1" as a part of the sense hat package for Raspberry Π compared to a high-quality IMU Xsens MTi-G 710. The aim of this research is the derivation of a stochastic error model for low-cost IMUs to emulate their characteristics. The model shall be sufficient to improve the accuracy of acquired data by signal processing to be used in real-time autonomous systems.