Physical AI Based on NPU Servers: UAE Water Purification Robot Solution

ECOPEACE's autonomous water pollution detection and purification robot ECOBOT has introduced Rebellions ATOM™ NPU servers into local infrastructure in the UAE. A central server analyzes video data collected by the robot to identify pollutants, and uses depth estimation to calculate 3D distances, supporting the robot's approach and recovery operations.

The background lies in the problem of wastewater and oil slicks generated by the oil and gas industry in the Middle East. Managing water bodies through direct human oversight is time-consuming, and if pollution is left unaddressed for a long period, additional treatment costs grow larger.

The core of the architecture is that precise AI inference based on high-resolution video is handled not by an edge device but by a central NPU server. This reduces the computational burden on the robot side, improves battery efficiency, and secures the possibility of operation with low-power, low-cost infrastructure.

The main operations are as follows.

• The ECOBOT at Dubai's Al Jadaf Port transmits video data
• The NPU server in Abu Dhabi performs object detection and depth estimation
• The robot, having received the analysis results, approaches the pollutant
• The oil recovery pump is activated to carry out suction and filtering treatment

Results have also been confirmed. Through TTA (Telecommunications Technology Association) certification, a 2x performance-per-watt level compared to GPU was officially verified, and NPU-based inference infrastructure was successfully built in the UAE. This case shows that even in a Physical AI environment where multiple AI models interact, Rebellions NPU can be a GPU replacement option.