Super-resolution Algorithms for Joint Range-Azimuth-Doppler Estimation in Automotive Radars

Super-resolution Algorithms for Joint Range-Azimuth-Doppler Estimation in Automotive Radars

Gürcan, Y.

Yarovoy, A. (mentor)

Electrical Engineering, Mathematics and Computer Science

Microelectronics

2017-01-25

In FMCW radar systems, spectral estimation methods are used to determine the beat frequency of the reflected signal in order to obtain information about the range of scatterers relative to the position of the transmit/receive elements. Array processing techniques take advantage of the phase shifts per element and likewise use spectral estimation methods to determine angle information. Range-Doppler processing methods utilize the Doppler shifts per chirp caused by the motion of the reflector to estimate the radial velocities relative to the position of the receiver. The problem of estimation of these three parameters of scatterers in an area can then be simplified as a problem of spectral estimation over three separate domains. The recent algorithms for solving such a problem in the context of array based radar applications range from adaptive, nonparametric methods, subspace based parametric methods and sparse recovery methods. However, these methods mostly ignore the inter- and intra- pulse Doppler effects of multiple moving targets. Furthermore, with demand for more resolution, investigation of wideband effects on range-azimuth-Doppler estimation is necessary. In this thesis, we firstly develop a MATLAB based MIMO radar simulator, on which the theoretical models are to be tested. The first aim of the thesis is to investigate the joint range-azimuth estimation methods and the shortcomings of narrowband assumptions, especially on the DoA estimation problem. We propose a novel signal model specifically for LFMCW radar systems and two methods of joint estimation using this model. The next aim is to investigate the effects of the movement of the targets on accuracy of the estimation problem. We study these effects firstly on the range-azimuth estimation problem and provide a more detailed signal model, which, theoretically also allows for the joint estimation of range, azimuth and velocity parameters using only a single chirp. Then we look into joint estimation in three domains and provide a subspace based algorithm, using the proposed signal model, capable of solving this problem. The accuracy of the new signal model and performance of the estimation algorithm is then tested on data generated by the MATLAB based simulator. Comparisons are made with respect to state of the art, MUSIC based joint estimation algorithm and theoretical bounds. Furthermore, range-Doppler estimation is performed on real life data, taken from the Dolphin prototype NXP automotive radar system. Future works based on this model is proposed.

http://resolver.tudelft.nl/uuid:4d4e1516-3c86-4b78-8c13-bf89953af029

Student theses

master thesis

(c) 2017 Gürcan, Y.