Ruggedized Ultrasound Sensing in Harsh Conditions: eRTIS in the wild

We present eRTIS, a rugged, embedded ultrasound sensing system for use in harsh industrial environments. The system features a broadband capacitive transducer and a 32-element MEMS microphone array capable of 2D and 3D beamforming. A modular hardware architecture separates sensing and processing tasks: a high-performance microcontroller handles excitation signal generation and data acquisition, while an NVIDIA Jetson module performs GPU-accelerated signal processing. eRTIS supports external synchronization via a custom controller that powers and coordinates up to six devices, either simultaneously or in a defined sequence. Additional synchronization options include bidirectional triggering and in-band signal injection. A sealed, anodized aluminum enclosure with passive cooling and IP-rated connectors ensures reliability in challenging conditions. Performance is demonstrated in three field scenarios: harbor mooring, off-road robotics, and autonomous navigation in cluttered environments, demonstrates that eRTIS provides robust sensing in situations where optical systems degrade.

💡 Research Summary

The paper presents eRTIS (Embedded Real‑Time Imaging Sonar), a rugged, fully embedded ultrasound sensing platform designed for harsh industrial environments where optical sensors suffer from dust, fog, rain, and other visual obstructions. The system combines a broadband capacitive transducer (SensComp 7000) driven by a high‑voltage bias and two‑stage amplifier with a 32‑element MEMS microphone array (SPH0641LU4H‑1). Two array geometries are supported: a regular 6 × 5 grid for spatial smoothing and a Poisson‑disc random layout that suppresses grating lobes and improves angular resolution across a wide frequency band (20 kHz–80 kHz).

The hardware architecture is split into a front‑end “transceiver” PCB and a back‑end “data acquisition” PCB. The front‑end handles excitation signal generation (log‑FM chirps) and reception circuitry. The back‑end is built around an STM32F429 microcontroller, which uses its internal DAC to synthesize the excitation waveform, generates the PDM clock for the microphones, and captures all 32 digital streams via DMA into a 64 MB external SDRAM buffer. Deterministic timing and phase‑coherent sampling are ensured, which is critical for accurate beamforming.

Data are transferred to an NVIDIA Jetson module (compatible with Nano, TX2, Xavier NX, Orin variants) over a high‑speed USB 2.0 link implemented with the STM32’s ULPI peripheral and an external USB3300 PHY. The Jetson provides a CUDA‑enabled GPU for computationally intensive tasks such as 2‑D and 3‑D digital beamforming, frequency‑domain filtering, envelope detection, and target localization. By off‑loading the heavy signal‑processing workload to the GPU, the system achieves sub‑30 ms end‑to‑end latency, enabling real‑time perception loops for autonomous robots.

External synchronization is realized through a custom controller that can power and trigger up to six eRTIS units simultaneously or in a defined sequence. The controller also supports bidirectional triggering and in‑band signal injection, allowing precise timing alignment across distributed sensor networks.

Mechanical robustness is achieved with a CNC‑machined anodized aluminum enclosure that meets IP65 (or higher) standards. The case serves as a passive heat sink for the Jetson module; thermal tests under sustained CPU and GPU load showed maximum temperatures below 66 °C after two hours of continuous operation, confirming adequate passive cooling without fans. All connectors (power, Ethernet, RS‑422, M12) are sealed against water and dust, and a 9‑axis IMU (ICM‑20948) provides orientation data for motion compensation.

The authors validate eRTIS in three real‑world scenarios: (1) harbor mooring vessel monitoring, where fog and mist are prevalent; (2) off‑road robotics in dusty, unstructured terrain; and (3) autonomous guided vehicle navigation in a cluttered indoor factory. In each case, the system maintained distance errors below 0.2 m and angular errors within ±5°, outperforming camera and LiDAR solutions that failed under the same conditions. The 3‑D microphone array enabled real‑time obstacle maps without mechanical scanning.

Key contributions include: (i) the first fully embedded, broadband, 32‑channel 3‑D in‑air ultrasound sensor tailored for industrial harshness; (ii) an heterogeneous STM32‑Jetson architecture that delivers low‑latency GPU‑accelerated processing at the edge; (iii) a modular, IP‑rated enclosure with passive thermal management; and (iv) a flexible external synchronization scheme supporting multi‑sensor deployments. The paper positions eRTIS as a viable alternative or complement to optical perception in demanding environments and outlines future work on higher‑frequency operation, power‑optimized designs, and distributed synchronization protocols.