IIIT-H, Ati Robotics demonstrate AI robot navigation breakthrough | Hyderabad News

Robot deployed at Ati Motors

Hyderabad: Research conducted over three years by the International Institute of Information Technology, Hyderabad (IIIT-H), in collaboration with Bengaluru-based Ati Robotics, has moved from the laboratory to a factory floor, where a robot demonstrated a new way of navigating its surroundings.The demonstration was made possible by research into vision-language models, which help computers interpret visual information, and image-based navigation. Recently, an autonomous mobile robot (AMR) navigated to a specific object on Ati Robotics’ factory floor using this technology, the institute said. “There was much cheer. Not so much for its assignment, because the task itself was rather simple. Our robots perform harder tasks in customer factories every day,” the company spokesperson said. The demonstration was significant, however, because of how the robot navigated its surroundings.The robot did not rely on LiDAR, a technology that uses laser beams to measure distances and map surroundings, or on a conventional map showing precise distances and positions. Instead, it used a camera, a series of images captured during an earlier walkthrough of the factory, and a 3D AI model that used pairs of images to work out the layout of the space. The spokesperson said the demonstration was the result of a long-standing partnership with IIIT-H’s Robotics Research Centre, which has been exploring navigation technologies that do not rely on LiDAR.“Humans do not navigate with 3D coordinates. Our sense organs detect our surroundings, and by combining visual cues, memory and a sense of movement, we figure out where we are. If we need to get to Point B from Point A, we use our knowledge of the destination and surroundings to decide how to go; we continuously adjust our plans, alter our route if a path is blocked, predict what others may do, swerve to avoid hitting someone and so on,” says Madhava Krishna, professor, IIIT-H, explaining that the research team tried out something similar for AMRs using vision-language models.