Abstract
Real-world situations require DSRC-based vehicular ad hoc networks that employ sensor fusion and perception systems. Direction of Arrival (DOA) estimation is critical for sensor fusion as it helps track moving objects. This paper presents a new approach to DOA estimation for autonomous vehicles, using a nested sensor array and the Cuckoo Search (CS) Algorithm. A non-uniform nested sensor array, situated on the vehicle, is proposed. The non-uniform array of sensors assists the estimation of DOA with improved angular resolution and is more robust to ambiguities. The CS Algorithm is implemented within the proposed methodology to optimize the iteration for DOA estimation with improved accuracy and speed related to convergence. A thorough comparison of the proposed nested sensor array to a classical sensor array is performed. An example of the CS Algorithm for DOA estimation is presented, along with some real world examples. We demonstrate the power of the algorithm to accurately estimate the direction of incoming signals, even in challenging environments with noise and interference and CS also performs extensive simulations and experimental validation by utilizing the highly non-linear fitness function. This research represents a tremendous advancement in the sensor fusion and DOA estimation for autonomous vehicle research. This presents a much more robust system for perception, aiding autonomous vehicles create a more thorough context when preventing collisions and assisting self driving vehicles aid in safer navigation. The proposed model investigates local and global minima of highly non-linear functions for DOA estimation, and assesses performance using the frequency distribution of RMSE, variability analysis of RMSE, estimation accuracy, RMSE of CDF, and robustness against snapshots and noise and RMSE for Monte Carlo simulation runs.