Design And Optimization Of Silicon-Based Optical Phased Array For LIDAR

Optical phased array(OPA)LIDAR is a sensing technology that uses laser pulses to measure the distance to an obstacle object.The traditional mechanical scanning radar is costly,large in size,power consumption,heavy in mass,and is susceptible to many instabilities at high rotational speeds,and the device needs to be calibrated frequently.In contrast,phased-array LIDAR overcomes these disadvantages well,and not only that,but also has the advantages of high sensitivity and efficient multi-target detection.Optical phased array LIDARs are expected to be widely used in the next decade for autonomous driving,robotics,imaging,unmanned aerial vehicle(UAVs),national security,health-care,Internet of Things and so on.In this paper,we make improvements based on two different types of silicon-based phased-array LIDARs and propose an improvement strategy for the calculation of the optical steering element-grating coupler,which is studied as follows:1.Silicon-based optical phased arrays have been widely explored,while the design of steering with high side-lobe suppression capability remains a great challenge.To this end,this paper carries out the optimization of an optical path differential phase modulated 3D OPA with Si3N4 as the core layer and SiO2 as the cladding layer.The results show that the Particle Swarm Optimization(PSO)algorithm can effectively suppress the side-lobes of the OPA substantially.After that,the effects of the Gaussian incident light source mode field diameter and the alignment error of the Gaussian light source in OPA on the OPA performance are investigated.The study in this paper shows that aperiodic OPA has more performance advantages than uniform OPA,at the same time,provides an effective way to optimize randomly distributed OPA within a controlled time and a large range of parameters.2.To study optical phased arrays based on non-uniform arrays of grating couplers,we use Genetic Algorithms(GA)to optimize the Side Mode Suppression Ratio(SMSR)in far-field pattern.In the OPA study,we focus on the optimization of the main-lobe position versus wavelength in a two-dimensional nonuniform array.In addition,the effect of the fabrication error of the array on the SMSR is also analyzed,and the results show that the position error has a greater effect on the SMSR.This study provides an effective way to optimize a large number of parameters for a randomly distributed OPA in a controlled time frame.3.For the first time,a subwavelength arc-blazed grating coupler is introduced into OPA as a single array element,and a larger far-field intensity was obtained.Usually,the design of subwavelength grating couplers requires a difficult modeling and electromagnetic simulation process,and the results are generally difficult to predict,due to the fact that the structure is complex and the parameter space is huge,making optimization more difficult.In this paper,we use a Deep Neural Network(DNN)combined with an Auto-Encoder(AE)network to construct a forward prediction network containing full connectivity to predict the far-field spectrum line by line and synthesize the far-field pattern,Based on this,this work introduces the inverse design idea to construct a Convolutional Neural Network(CNN),which can effectively predict the parameter combinations corresponding to the far-field patterns.The results show that the two neural networks have 99.13%and 99.79%accuracy on the test set,and at the same time,the speed of electromagnetic simulation of the arc-blazed grating coupler is reduced to the order of ms,which is more than 60,000 times faster than that of electromagnetic simulation.