Range-Adaptive FMCW Radar: Real-Time Adaptation of RF and Baseband for Robust and Energy-Efficient Sensing

Static radar parameters fix the obtainable sensing range and force high RF power/gain, often wasting energy and requiring multiple modules to cover different ranges. We develop a closed-loop, range-aware adaptation framework that tunes RF front-end and ADC parameters on a per-frame basis from the detected range, thereby matching the detectable distance to the scene while dynamically scaling RF front-end performance. A lightweight adaptation pipeline refines detections, applies temporal smoothing, and maps the next-frame configuration using a pre-trained ML model under an energy-cost constraint. Implemented in closed loop on a commercial radar module, the approach reduces total energy by 33 to 69% without degrading detection performance. The end-to-end latency including signal processing and configuration upload, achieved a median of 21.76ms, validating real-time operation.