Reduced Latency and Power Requirements in Automatic Targeting Algorithms using Meta-optic Elements

OBJECTIVE: Investigate and develop meta-optic elements that perform mathematical transformations to reduce latency and power requirements using artificial intelligence algorithms for autonomous target recognition. DESCRIPTION: Computer vision-aided Autonomous Target Recognition (ATR) is an important DoD modernization priority that helps extend and enhance human capabilities. Current technologies employ advanced artificial intelligence and machine learning (ML) algorithms for feature generation and object classification. In pursuit of increased accuracy, state-of-the-art models have become increasingly complex, thereby introducing latency into the system. For many applications, some degree of latency is acceptable. However, there is an emerging set of operations where latency reduction is paramount as the decision-making process is currently too slow for autonomous systems to respond appropriately. This is especially true when dealing with complex environments and extremely fast-moving objects such as hypersonic vehicles. One solution for reducing latency is to offload some of the mathematical operations from the processor to the optics. For example, optical analog computation can be used to perform Fourier filtering for image contour detection. This operation produces an image directly on the focal plane that is composed only of contours, effectively performing an image differentiation calculation at the speed of light without any computational expense. In general, optical analog computation performs pre-processing steps that compress the image data and enable ML model solutions with fewer coefficients and lower latency compared with models used on raw data. One drawback of conventional optical analog systems is that they require bulky lenses and large volumes that are not suitable for SWAP-constrained platforms. A promising option to replace conventional optical elements and reduce SWAP is to employ ultra-thin meta-optic devices. Meta-optics, in contrast to bulky refractive optics, produce customized optical functions using subwavelength scatterers that provide abrupt spatially varying phase shifts. By engineering the geometry and arrangement of scatterers, meta-optics can be designed to perform complex mathematical transforms on the incoming light. When applied to ATR algorithms, these meta-optic operators could provide orders of magnitude reduction in latency while maintaining low size, weight, and power requirements for ATR platforms. In this program, the Army seeks to partner with a small business and university participant to investigate and develop meta-optic elements for optical analog computation. These elements will perform the computationally expensive operations of ATR algorithms, leading to significantly reduced latency and power requirements. Results from this effort would support a number of DoD programs and general performance objectives for the optical element include low SWAP, large bandwidth, low f-number, and high efficiency. PHASE I: Design a thin (