We conduct research aimed at developing cutting-edge AI technologies and applying robot AI in the field robotics to enhance robot perception, intelligence, and automation capabilities. Our research spans across all areas of robotics, including robot AI for camera vision, sonar, active sensors, and control strategies.
AI research themes include enhancing underwater ultrasonic camera perception, reinforcement learning for underwater robot motion enhancement, and underwater object recognition and retrieval.
We have developed a method to synthesize realistic sonar images using a Generative Adversarial Network (GAN). A ray-tracing-based sonar simulator calculates semantic information of a viewed scene, and the GAN-based style transfer algorithm then generates realistic sonar images from the simulated images. [Link]
We developed the detection and removal of crosstalk noise using a convolutional neural network in the images of forward scan sonar. Also, we applied the method to a three-dimensional point cloud generation and generated a more accurate point cloud. [Link]
This study explores the fusion of LiDAR and sonar sensors for mapping purposes in both surface and underwater environments. By acquiring data with LiDAR sensors on the surface and sonar sensors underwater, the goal is to improve mapping accuracy and coverage across both domains through integrated sensor data.
Underwater sonar and vision-based 3D shape reconstruction methods. The methods are used to localize and recognize underwater objects and generate large-scale underwater maps. [Link]