The forgotten spectrum: Reviving ultrasound for robust autonomy

Area of Science:

• Robotics
• Artificial Intelligence
• Sensor Fusion

Background:

• Autonomous systems heavily rely on visual sensors.
• Environmental conditions like fog, dust, or low light can impair vision.
• Alternative sensing modalities are crucial for robust operation.

Purpose of the Study:

• To investigate the efficacy of ultrasound in autonomous navigation.
• To enhance ultrasound performance using edge AI denoising.
• To improve system robustness when visual perception fails.

Main Methods:

• Utilized a low-cost ultrasound sensor array.
• Implemented an edge AI model for real-time noise reduction on ultrasound data.
• Tested autonomous navigation scenarios with simulated visual occlusions.

Main Results:

• Edge AI denoising dramatically improved ultrasound image clarity.
• Autonomous navigation accuracy increased by 30% in degraded visual conditions.
• Ultrasound provided reliable depth and obstacle data when vision was unavailable.

Conclusions:

• Combining ultrasound with edge AI denoising offers a cost-effective solution for enhancing autonomous system resilience.
• This approach provides a viable fallback mechanism for perception systems.
• It enables more dependable operation in challenging, real-world environments.