Oil spills caused by extraction, transportation, refueling, and equipment failures are a major threat to marine ecosystems. Once released, oil spreads rapidly due to wind, waves, and currents, making containment difficult and increasing environmental damage. In 2024, the International Tanker Owners Pollution Federation Limited (ITOPF) reported six large oil spills, releasing approximately 700 tons of oil. To address these challenges, an autonomous oil spill containment system is proposed.

The proposed system consists of three major subsystems: Sensing stage, path planning stage, and the actuation stage. Upon the detection of an oil spill via either the ship's on board sensors or through manual detection, the Underwater Oil Containment Vehicle (UOCV) gets activated. The UOCV gets deployed from the ship onto the water surface. The UOCV has a deflated boom attached to it on one end for oil spill containment, having the capability to get inflated via an onboard air compressor.

As soon as the UOCV hits the water surface, the first subsystem gets activated. The sensing subsystem comprises a stereo camera for both the oil spill area and distance detection. A CNN (Convolutional Neural Network) based computer vision algorithm has been developed for real-time detection of the oil spill area. The algorithm has been trained on a series of oil spill images at various brightness levels, circumferential areas, oil types, and proximities. The first instance of the oil spill detection acts as a trigger variable for the path planning stage to kick in.

A reactive edge-following path planning algorithm is developed. The primary objective of using such an algorithm is its ability to adjust to the variable area of the oil spill in real time. As soon as the initial circumferential boundary of the oil spill has been estimated, directional vector fields are computed along the boundary to determine the waypoints for the UOCV to follow. The algorithm is continuously updated in real time based on boundary updates provided by the CNN framework. Once the waypoints have been determined, a conventional PID algorithm actuates and controls the motor-propeller onboard the vehicle. After the UOCV bounds the oil spill area, the boom is inflated to contain the oil spill. This approach can also be extended to incorporate booms with adsorbent coatings for oil adsorption. The proposed framework ensures minimal human intervention and rapid response, preventing the oil spill from spreading across the water surface.

Ultimately, the UOCV reimagines the oil spill containment as an autonomous, real-time process rather than a reactive, manual-driven one. What we are demonstrating is a solution that is not only technically feasible but faster, smarter, and more scalable than what the current market provides, with the end goal being to take a meaningful step towards protecting our waters and environment with minimal human intervention.