Algorithmic Enhancement of Automotive MEMS Gyroscopes With Consumer-Type Redundancy

In this paper a novel fusion algorithm for a setup of redundant MEMS gyroscopes is proposed. Six cost-efficient, tri-axial, consumer-type sensors are combined with three single-axis vibration-robust automotive-type devices. The proposed algorithm provides a virtual sensor output that has both low noise due to redundancy averaging, as well as good mitigation of external, high-frequency vibration disturbances. Thereby the advantages of both sensor types, consumer and automotive, are combined. If one sensor is compromised by vibration influences, the conflicting sensor signals are challenging to conventional fusion methods. To overcome this, a two-stage fusion algorithm, called topological conflict measure (TCM), is proposed. It is based on weighted sums, where the weights for each sensor diminish as the signal reliability decreases. The method can be utilized to (1) assess a measure of multi-sensor conflict, (2) to improve fusion of redundant devices by lessening weighting for sensors with high conflict, and (3) to combine the fused signals calculated from sensors with inherently different characteristics, like vibration robustness in this case. Furthermore, a selection of other possible methods is implemented and compared to TCM: simple Bayesian averaging (BA), Dempster-Shafer Evidence Theory (DSET), Kalman filtering (KF) and fuzzy inference system (FIS). All methods offer similar, improved noise levels while the TCM, KF and FIS methods achieve the best vibration rejection. Among those, TCM has the best computational efficiency and adaptability to other types of redundant sensor setups.