Embedded Air-Coupled Ultrasonic 3D Sonar System with GPU Acceleration

Air-coupled ultrasonic sonar systems are a valuable complement to lidar sensors for safe navigation of autonomous vehicles, especially in environments with optically transparent and reflective surfaces. Consequently, sonar systems based on phased arrays and beamforming are a promising compact solution, providing precise and unambiguous 3D localization of multiple objects in a wide field of view. However, a cost-effective and viable integration is challenging due to the increased component count and computational effort. Therefore, we present an embedded phased array sonar system requiring a minimum number of low-cost components, and providing high frame rates. The system contains 36 digital MEMS microphones, one 40 kHz transducer, an FPGA for CIC-filtering and communication with the main processor (Nvidia Jetson Nano) via Hi-Speed USB or WiFi. The GPU accelerated frequency-domain signal processing consisting of matched filtering, optimized conventional narrow-band beamforming and envelope extraction, enables frame rates up to 30 Hz. In an anechoic chamber, a field of view of ±80° and an angular resolution of 14° have been measured using two O100 mm spheres. Therefore, this self-contained phased array sonar allows a practicable sensor fusion with lidar data for a more reliable monitoring of the environment.