Optimized Autonomous Avoidance for Underwater Compact Towed Vehicles

As underwater towed vehicles (UTV s) are increasingly used for underwater data collection and coral reef monitoring, the need for effective underwater obstacle avoidance solutions has become more pressing. Yet, the underwater en-vironment poses unique challenges for above-water equipment, while underwater equipment is significantly more expensive. For instance, stereo cameras and depth sensors, which are commonly used for obstacle avoidance in terrestrial and aerial domains, are difficult and costly to deploy and maintain underwater due to the water attenuation, refraction, and corrosion effects. In this paper, we present an underwater-specific visual obstacle avoidance system that runs on a low-energy chip onboard the vehicle at high processing speeds. The proposed system is both cost- and space-efficient, making it ideal for smaller towed vehicles. A comparison to existing methods is provided to illustrate processing speed and accuracy.