Interferometric Microwave Radar With a Feedforward Neural Network for Vehicle Speed-Over-Ground Estimation

A novel approach to measuring the ground speed of autonomous vehicles using interferometric radar is presented. Using a microwave radar with interferometric processing and a feedforward neural network with local regression, high-accuracy velocity estimation is achieved with a downward-facing radar that, unlike forward-looking Doppler radars, can be protected from road debris and is, furthermore, unaffected by wheel slip and requires no external inputs, such as Global Navigation Satellite Systems (GNSS). A 16.9-GHz active interferometric array generates a grating lobe pattern and as the ground passes through the pattern, the range of frequencies over which the response is distributed increases proportionally with the ground velocity. We implement a feedforward neural network to estimate the velocity based on the interferometer's frequency response. The estimator achieved a root-mean-squared error of 0.138 (m/s), equivalent to that of the forward-looking Doppler radars.