Multiple IMU Integration for Vehicular Navigation

One of the most widespread applications of the Global Positioning System (GPS) is vehicular navigation. Improving the navigation accuracy continues to be a focus of research, commonly answered by the use of additional sensors. A sensor commonly fused with GPS is the inertial measurement unit (IMU). Due to the fact that the requirements of commercial systems are low cost, small size, and power conservative, micro-electro mechanical sensors (MEMS) IMUs are used. They provide navigation capability even in the absence of GPS signals or in the presence of high multipath or jamming. This paper addresses a centralized filter construction whereby navigation solutions from multiple IMUs are fused together to improve accuracy in GPS degraded areas. The proposed filter is a collection of several single IMU block filters. Each block filter is a 21 state IMU filter. Because each block filter estimates position, velocity and attitude, the system can utilize relative updates between the IMUs. These relative updates provide a method of reducing the position drift in the absence of GPS observations. The proposed filter’s performance is analyzed as a function of the number of IMUs used and relative update type, using a data set consisting of GPS outages, urban canyons and residential open sky conditions. While the use of additional IMUs (including a single IMU) provides negligible improvement in open sky conditions (where GPS alone is sufficient), the use of two, three, four and five IMUs provided a horizontal position improvement of 25 %, 29 %, 32 %, and 34 %, respectively, when GPS observations are removed for 30 seconds. Similarly, the velocity RMS improved by 25 %, 31%, 33%, and 34% for two, three, four and five IMUs, respectively. Attitude estimation also improves significantly ranging from 30 % – 76 %. Results also indicate that the use of more IMUs provides the system with better multipath rejection and performance in urban canyons.