Toward Data-Driven STAP Radar

TL;DR

STAP radar performance is fundamentally limited by unknown clutter statistics, leakage between target and clutter subspaces, and heavy computation for large-dimensional adaptive processing. The paper proposes a data-driven workflow that uses RFView to generate heatmap tensors of MVDR beamformer outputs and leverages computer-vision–inspired methods to detect and localize targets, with the concept of a COCO-like STAP radar database. A regression CNN is developed to estimate target location from heatmaps and shows substantial improvements over the conventional MVDR peak method, achieving up to sevenfold reductions in localization error on a sizeable RFView-based dataset. This approach demonstrates a scalable, realistic pathway toward learned STAP radar processing and opens questions about generalization, sensor configurations, and the speed-accuracy trade-offs in data-driven radar systems.

Abstract

Using an amalgamation of techniques from classical radar, computer vision, and deep learning, we characterize our ongoing data-driven approach to space-time adaptive processing (STAP) radar. We generate a rich example dataset of received radar signals by randomly placing targets of variable strengths in a predetermined region using RFView, a site-specific radio frequency modeling and simulation tool developed by ISL Inc. For each data sample within this region, we generate heatmap tensors in range, azimuth, and elevation of the output power of a minimum variance distortionless response (MVDR) beamformer, which can be replaced with a desired test statistic. These heatmap tensors can be thought of as stacked images, and in an airborne scenario, the moving radar creates a sequence of these time-indexed image stacks, resembling a video. Our goal is to use these images and videos to detect targets and estimate their locations, a procedure reminiscent of computer vision algorithms for object detection$-$namely, the Faster Region-Based Convolutional Neural Network (Faster R-CNN). The Faster R-CNN consists of a proposal generating network for determining regions of interest (ROI), a regression network for positioning anchor boxes around targets, and an object classification algorithm; it is developed and optimized for natural images. Our ongoing research will develop analogous tools for heatmap images of radar data. In this regard, we will generate a large, representative adaptive radar signal processing database for training and testing, analogous in spirit to the COCO dataset for natural images. As a preliminary example, we present a regression network in this paper for estimating target locations to demonstrate the feasibility of and significant improvements provided by our data-driven approach.

Figures (4)

• Figure 1: Representative $5\times26\times21$ heatmap tensor from the RFView example scenario (Section \ref{['Sec4']}). Each heatmap pertains to a unique range bin.
• Figure 2: The map of the area of study. The blue triangle indicates the platform location and the red region is the constrained area for target placement. The elevation heatmap overlaying the left image represents the simulation region. Left image source: USGS EarthExplorer.
• Figure 3: CNN architecture used for the regression network.
• Figure 4: Comparing average absolute error for target localization using the regression CNN model (in blue) and the traditional peak-finding method with MVDR beamforming (in red), for an average SCNR of $-2.82 \ \text{dB}$.