FPGA-based implementation of Kalman filter for real-time estimation of tire velocity and acceleration

Abstract

Antilock braking systems require accurate vehicle speed and acceleration estimations from wheel speed sensors. Adaptive Kalman filtering is used to estimate wheel speed and acceleration for each wheel. Single core or multicore processor-based embedded systems perform the required estimation using a two-stage extended Kalman filter in many clock cycles, which limits Kalman filter speed and accuracy. In the proposed system, purely arithmetic operations used in the process are simplified and implemented as embedded logic to reduce the overhead delays and increase the update rate by 48%. The proposed system is run sequentially using finite-state machines, while a large number of multiplication and division operations are removed through rewiring and logic reuse. The simplifications not only reduce the amount of logic required to perform the operations by 39.7% but also use simple logic, which can run faster than the replaced logic on field-programmable gate array. The results presented show comparable root-mean-square wheel speed error with a high update rate of 20 MHz when running on a 50-MHz board in all scenarios. The system also provides accurate wheel acceleration estimation, which is crucial for many systems, including the reference system.