Time-Frequency Range Estimation Method for Pulsed LiDAR

Abstract

Light detection and ranging (LiDAR) is the key sensor behind the 3D perception of self-driving vehicles. The range, precision and accuracy of a LiDAR sensor mainly depend on how each received pulse is captured, sampled and processed. However, to the best of our knowledge, such a topic is yet to be investigated in depth. This work aims to fill such a gap by introducing a new range estimation method. The method takes only 3 samples from the rising edge of the received pulse and combines two simple estimation techniques: time-domain edge detection and frequency-domain phase estimation. The proposed method was demonstrated over a dynamic range of 45 dB, proving to be robust over the entire interval. Sub-centimeter precision and accuracy were achieved over a dynamic range of 24 dB. The obtained estimations are bounded by a derived theoretical limit that is identical to a Cramér-Rao lower bound of a phase estimator.