In today's automotive applications, radar is widely used to estimate the target position and velocity with respect to the radar position. Estimating the position of the target in terms of range and velocity is fairly advanced and accurate. However, resolving two closely spaced targets in the azimuth domain which have the same range and relative radial velocity as seen from the radar is still a very challenging problem that needs to be addressed. The azimuth resolution that can be achieved by a radar sensor is directly proportional to the number of antenna elements in the radar, that is increasing the number of elements improves the resolution capability. But increasing the aperture of the single radar mounted on the automobile is not desirable as it is cost ineffective and not easy to fit in the design of the automobile. Thus a study is performed on combining the data from multiple radar sensors which are distributed over the fascia of the car along the horizontal axis. In this thesis, methods to achieve higher angular resolution in azimuth using multiple radar sensors is presented.

The system design and geometry for a distributed system consisting of FMCW radar sub-systems is studied along with the signal model for the same. The interpretation of near field and far field region around the radar sensors is presented. In this thesis, the Range-Doppler processing of the data is performed first and the snapshot related to the range and Doppler bin where the target is detected is extracted to perform the Direction of Arrival (DOA) estimation. This is done as it is computationally efficient and the DOA estimation can be performed on every snapshot and hence the changes in the target position can be detected faster. The signal model for DOA estimation using single sensor and distributed sensors is provided for single snapshot case. Sparse signal processing technique is chosen to estimate the DOA in this thesis as it has better performance and certain advantages in DOA estimation of target using single snapshot compared to other methods like Beam forming or MUSIC. The theory behind compressive sensing technique is discussed along with the concept of block sparsity to fuse the data from multiple sensors. A new algorithm called Block Focal Under determined System Solver (FOCUSS) is proposed to incoherently combine the data from multiple sensors in order to achieve a better performance. This method benefits from the spatial diversity gain by combining the data from multiple sensors. However, the resolution improvement achieved by the incoherent combining of the data from multiple sensors is still limited by the largest aperture of the sub-system used and hence a way to coherently combine the data from multiple sensors is presented. Coherent FOCUSS algorithm is proposed which combines the data from multiple sensors coherently and the virtual aperture of such combining is given by the separation between the sensors and hence it can achieve very high resolution in azimuth. Such a method has some drawbacks when the target is non-isotropic or the distance between the sensors is too large, which is also discussed in this thesis. A method called Fusion FOCUSS is introduced to overcome some of the drawbacks of coherent processing, whose resolution capability is in between that of Block FOCUSS and Coherent FOCUSS.

Simulations are performed to evaluate the performance of the proposed algorithms and for comparison purpose the results obtained from Block Orthogonal Matching Pursuit (BOMP) algorithm is provided. Monte Carlo runs are performed for different scenarios consisting of varying SNR, target phase and baseline of the distributed system. The performance of the algorithms with varying SNR values is presented. The penalty incurred in performing coherent processing on a non-isotropic target is discussed. The problem of off-grid targets is studied and a possible solution for the same is implemented as discussed in literature. Results obtained by performing an experimental evaluation in the anechoic chamber to study the performance of Block FOCUSS is presented along with the explanation of results. We also propose some ideas for future work to further investigate the problem.