UAV-SAR: Simultaneous Radar-Based Odometry and Synthetic-Array Sensing for Unmanned Aerial Vehicles
David Hunt, Shaocheng Luo, Samuel Rivera, Aarav Prakash, Cameron Morris, Tingjun Chen, Miroslav Pajic
AI summary
Problem
Traditional UAV navigation relies on GPS, cameras, or lidar, which fail in GPS-denied or visually degraded environments. Prior radar-based solutions address odometry and high-resolution sensing separately due to conflicting temporal requirements, leaving a critical gap for unified, all-weather UAV perception.
Approach
The system fuses downward- and outward-facing mmWave radar data with an IMU via an extended Kalman filter for robust odometry, then leverages this motion data to coherently integrate multiple radar frames into a dynamic synthetic array, finally applying a lightweight U-Net model to convert the beamformed responses into high-resolution point clouds.
Key results
- Unified architecture for simultaneous radar-based odometry and synthetic-array sensing
- Angular resolution improved by an order of magnitude via dynamic synthetic array beamforming
- Lightweight U-Net model generates dense point clouds using 310× fewer training samples than prior work
- Real-time ROS 2 and PX4 autopilot integration demonstrated stable indoor flight and reliable sensing
Why it matters
Enables reliable UAV operation in GPS-denied and visually degraded environments where traditional sensors fail, advancing autonomous navigation for defense, search-and-rescue, and logistics.
Abstract
Unmanned aerial vehicles (UAVs) require accurate odometry—i.e., estimating the position and velocity of the vehicle over time—as well as high-resolution sensing to safely and effectively operate in complex environments. Traditionally, GPS, cameras, and/or lidar sensors have been used to perform these functions. However, GPS can be jammed in contested environments while cameras and lidars fail in visually degraded conditions, limiting UAV operations in these scenarios. In this work, we present UAV-SAR, a unified architecture that utilizes mmWave radars to simultaneously achieve precise odome- try measurements and perform high-resolution synthetic-array sensing. Here, UAV-SAR measures a UAV’s altitude and velocity from downward- and outward-facing radars and fuses these measurements within a commercially available flight controller to produce accurate odometry estimates. These odometry esti- mates are then used to dynamically construct synthetic arrays by coherently integrating multiple radar frames together over a duration of 0.5 s, improving the angular resolution by an order of magnitude compared to the physical array alone. Finally, a lightweight deep learning model is utilized to convert high- resolution range-angle responses into 2D point clouds suitable for downstream perception tasks. UAV-SAR is validated on a custom UAV prototype where it is integrated with ROS2 and the PX4 autopilot to demonstrate stable flight, reliable odometry, and high-resolution radar sensing in indoor environments.