Scholarly record

TESTING OF A SMART ALGORITHM FOR GPS DATA PREDICTION TO FUSE THEM WITH THE HIGH RATE INERTIAL DATA IN A QUATERNIONIC MEMS-INS/GPS INTEGRATED NAVIGATOR

Teodor Lucian Grigorie, Costinel Laurentiu Corcau

Abstract

The paper presents a prediction algorithm for the low rate GPS data in order to fuse them with the high rate Inertial Navigation System (INS) data; a major shortcoming in the integration process of the two navigation systems is due to the significant difference in theirs data rates in the reading processes. The algorithm is based on a smart mechanism based on adaptive neuro-fuzzy inference systems (ANFIS) concept, and includes an adaptive network of fuzzy inference systems. Under the form of a five layer feed-forward network with adaptive and fixed nodes, the ANFIS contains a set of adaptive parameters characterising all the adaptive nodes, which are updated during a training procedure based on training data set. The learning procedure used by ANFIS is a hybrid learning algorithm combining the gradient descent (back propagation) method and the least square method (LSM). The proposed prediction algorithm is actually an extrapolator for the data read from the GPS system at a low rate, which uses at the first step some data from the MEMS-INS navigator to be created and initially trained. After this first step the prediction algorithm operates in the already trained configuration and estimates the samples that are lack from GPS signal for the times synchronized with the MEMS-INS times up. The structure runs and it is kept unchanged until to the time point when the GPS data are received again. Using the currently received GPS valid data and the earlier GPS readings as training data set, the structure is trained again and starting from this time point is used in this new configuration and so on.

Publication Impact Profile

• Captures
• Mendeley - Readers: 3

PLUMX see details

DOI resolver: doi.org/10.5593/sgem2017/22/s09.074

Publication details

Back to publication list