ABSTRACT
Kalman filter is one of the most powerful engineering instruments for variables estimation. Developed in 1960[1], this algorithm is widely used in dierent areas of technology. Nowadays there are many variations of this filter. All of them require calculation of dierent matrix operations and have quite high demands for computation resources. Since most DSP operate only on scalar (not matrix) variables, real-time performance, that is strongly needed in motor control applications, is dicult to achieve with cost-eective chips. FPGA-based Kalman filter implementations that use soft-processors have the same problem. Like DSP they do not use all of FPGA advantages such as parallel mathematical and memory access operations. This article presents a new FPGA-based coprocessor, that directly operates with matrix variables and is designed specially for Kalman filtering in motor control applications.
