AI-Controlled Robotics Autonomous Underwater Robot Collects Garbage from the Seabed

Related Vendors

FAULHABER Drive Systems

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

Wuxi InfiMotion Technology Co., Ltd.

94% of marine debris lies on the ocean floor. However, conventional collection methods fail. Researchers at TUM have developed an autonomous diving robot that independently collects garbage from the seabed.

The world's oceans today contain 29 to 73 million short tons of waste, with approximately 94% located on the seabed. So far, cleanup efforts have mainly focused on surface waste, while underwater recoveries usually require dangerous diving missions.

A research team from TU Munich (Germany) has now developed an autonomous diving robot within the EU project SeaClear that can independently identify and recover marine debris using AI-based object recognition and precise gripping technology. The system combines computer vision, ultrasonic technology, and adaptive force control into a fully integrated application.

Multi-Layered System Architecture With Distributed Intelligence

The SeaClear system operates based on a hierarchical approach: an unmanned service boat roughly maps the seabed using ultrasound and supplies the underwater units with power and data connection via cables. A 20 inches-long search robot efficiently scans large areas before the TUM diving robot, powered by eight mini turbines, navigates precisely to identified debris locations.

Object identification poses special challenges for image processing: since scarcely any training data for underwater scenarios exists, the researchers developed a dataset with over 7,000 labeled images. The AI analyzes camera footage and sonar data in parallel and generates 3D models of the target objects.

"Sonar enables orientation even in murky water, while the cameras provide detailed analysis," explains Dr. Stefan Sosnowski from the Chair of Information-Oriented Control.

Adaptive Gripping Technology With Force Feedback

The four-finger manipulator of the one-cubic-meter underwater robot combines high gripping force with precise sensors: at a maximum force of 4,000 Newtons, it can handle objects weighing up to 550 pounds, while integrated force sensors enable damage-free manipulation of even fragile items. The gripper control uses 3D data to optimize grip points, thus preventing object damage caused by improper force distribution.

Despite autonomous navigation, the 265 pounds system maintains a cable connection to the surface. This is a deliberate trade-off between mobility and operating time. The cable supply allows for continuous operation without battery changes and higher AI performance through external computing power. Buoyancy foam keeps the submarine in a neutral hovering state, enabling precise positioning during gripping operations.

Cost Efficiency from 52 feet Water Depth

Sosnowski's cost-benefit analysis shows: from a water depth of 52 feet, the autonomous system economically surpasses conventional diving operations. In Dubrovnik, the researchers documented over 1,000 pieces of debris in 100 square meters—a workload that highlights the capabilities of autonomous systems.

The SeaClear project successfully demonstrates the integration of various robotics disciplines: from AI-supported perception to adaptive manipulation and multi-robot coordination. For maritime robotics, the system opens up diverse fields of application beyond waste collection: from infrastructure inspection to underwater archaeology. (heh)

Machine Authentication

Machine Operator Authentication - Why knowing "who" matters