DKFI Leverages AI to Improve Environmental Awareness of Robot Underwater Vehicles

Credits: LeoWolfert / Shutterstock

concept: The German Center for Artificial Intelligence Research (DFKI) has developed a project called DeeperSense that combines visual and acoustic sensors with AI to improve the environmental awareness of robots’ underwater vehicles. The project aims to raise awareness of unmanned underwater vehicles (UUVs) in three use cases: diver monitoring in muddy waters, seafloor mapping, and coral reef exploration.

Nature of confusion: The DeeperSense project is based on the concept of intersensory learning, where one sensor modality learns from another. Thus, the output of one sensor is similar to the output of other sensors in terms of accuracy, output type, and data interpretation. In the case of UUV, it has a camera and sonar as two sensors that observe the same scene at the same time. Low resolution sonar data serves as input to the artificial neural network, and high resolution camera data serves as output. This combination gradually adapts to the network to provide the desired output and learns about the relationship between input and output data. The result is an algorithm that, when training is complete, produces a camera-like image based solely on low-resolution sonar data.

Outlook: One of the common maritime use cases DeeperSense has previously wanted to address was poor environmental awareness of UUVs due to muddy water, tight spaces, or dark areas. For diver monitoring in muddy waters, traditional monitoring systems are limited by the range of underwater optical sensors. The project introduced by DFKI trains UUV sensors to provide camera-like images that can be easily interpreted by humans at the control station. In the second use case of coral reef exploration, the challenge is reliable obstacle detection that can be overcome by a combination of visual and acoustic sensors. The AI ​​algorithm recognizes an object identified from one sensor’s data to another sensor’s data. In this way, instead of crossing the reef, you go through the reef. In the third use case, the project UUV reliably maps the seabed, which was expensive on the ship. The mapping done under the project is cheaper, more reliable, and provides detailed output. The application can be extended to exploration activities. The DFKI project has been awarded $ 3.5 million by the EU under the 2020 Research Framework Program.

This article was originally published DKFI Leverages AI to Improve Environmental Awareness of Robot Underwater Vehicles

Back to top button