@article { author = {Keighobadi, J and Yazdanpanah, M J and Kabganian, M}, title = {Fuzzy Variable-Length Sliding Window Blockwise Least Square Algorithm with Application to Vehicle Heading Determination}, journal = {Iranian Journal of Mechanical Engineering Transactions of the ISME}, volume = {9}, number = {2}, pages = {59-79}, year = {2020}, publisher = {Iranian Society of Mechanical Engineering}, issn = {1605-9727}, eissn = {}, doi = {}, abstract = {In ground vehicles, three-axis magnetometers may be corrupted by both soft- and hard-iron disturbances. Therefore, it may not be possible to achieve qualified headings without online calibration of this magnetic system. First contribution of this paper is focused on improving the order of persistent excitation of the squared signal matrix through incorporation of a direction cosine matrix in estimation model. As the main contribution, a fuzzy change detection scheme for adjusting the length of data sliding window of blockwise least square (BLS) algorithms is presented in the framework of on-line estimation of system parameters under both abrupt and gradual changes. This is called fuzzy variable-length sliding window (FVLSW) BLS. Two change detection indices including generalized likelihood ratio and averaged parameter estimation errors together with their changes are considered as inputs of the fuzzy system. The defuzzified outputs consists a forgetting factor in order to place more emphasis on the recent data, and two adjusted lengths of data history windows.Simulations and real experiments revealed that the proposed approach has superior performance with respect to the latest variable-length sliding window (VLSW) BLS estimation algorithm. The superiority is more significant when the measurement noise power is substantial.}, keywords = {fuzzy change detection,persistent excitation,vehicle navigation,least square,online calibration}, url = {https://jmee.isme.ir/article_45996.html}, eprint = {https://jmee.isme.ir/article_45996_56b9f7b18d40ac18e49b96906e09b296.pdf} }