Abstract:A pulse Doppler radar based real-time speed measurement system was designed by using CAN bus communication technology and multithreading technology according to the low speed characteristic of intelligent agricultural vehicle. In order to obtain a high precision result with the least system resource consumption, the multiresolution adaptive policy was designed by dividing speed level and variable sliding-window gray prediction. Experiments showed that the average error was 1.45% and the working frequency of resolution generator was reduced by 76.7% after adaptive policy started when the speed of vehicle was 0.2 m/s. The average error was 0.61% and the working frequency of resolution generator was reduced by 90.0% after policy started when the speed of vehicle was 0.8 m/s. The output frequency of system was steady 9.6 Hz and the frequency of resolution generator was dramatically reduced during variable speed movement.