Design optimization of a 77 GHz antenna array using machine learning

In this work, a 77 GHz patch antenna array is presented as a front-end communication tool for Advanced Driving-Assistance System (ADAS). A regression based mathematical modelling algorithm Modified Extensible Lattice Sequence (Mels) is used to develop a supervised machine learning based surrogate mathematical model from the physical array model with very few simulation iterations. Global response search method (GRSM) optimization routine is used for optimizing the surrogate model under a very short time to achieve the design optimization goals. A comparison is shown between traditional GRSM array optimization and optimization coupled with machine learning. Additionally, a simulation-based case study is presented using the optimized antenna array for vehicle to vehicle communication in a time-varying autonomous driving scenario.