IIIT-H, Ati Robotics advance vision-based AI navigation for autonomous robots

HYDERABAD: A three-year research collaboration between the International Institute of Information Technology Hyderabad (IIIT-H) and Bengaluru-based Ati Robotics has moved from the laboratory to a factory floor. An autonomous mobile robot (AMR) recently demonstrated a new approach to navigation, as reported by TOI. The robot navigated to a specific object inside Ati Robotics’ factory using […] HYDERABAD: A three-year research collaboration between the International Institute of Information Technology Hyderabad (IIIT-H) and Bengaluru-based Ati Robotics has moved from the laboratory to a factory floor. An autonomous mobile robot (AMR) recently demonstrated a new approach to navigation, as reported by TOI. The robot navigated to a specific object inside Ati Robotics’ factory using artificial intelligence (AI) technology. According to IIIT-H, the system combines vision-language models with image-based navigation. The demonstration showed how a robot can navigate using visual information and a model of its surroundings, without relying on light detection and ranging (LiDAR) technology. Robot uses images instead of LiDAR to navigate LiDAR uses laser beams to measure distances and map surroundings. The robot instead relied on a camera and images captured during an earlier walkthrough of the factory. Researchers also used a 3D AI model to help the robot understand the layout. The model analysed pairs of images to build a representation of the space and guide the robot towards its target. “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,” a company spokesperson said. However, the demonstration marked an important step because of the navigation method, the spokesperson added. Researchers draw on human visual perception Madhava Krishna, a professor at IIIT-H, said the team aims to replicate aspects of human navigation in autonomous mobile robots using vision-language models. “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,” he said. Krishna explained that humans use their knowledge of a destination and its surroundings to plan routes. They also adjust their plans when obstacles appear or circumstances change. The same principle guides the research team's approach. Robots must interpret visual cues, use contextual information and adapt their routes to navigate effectively. The factory demonstration highlighted the team's efforts to develop robots that can understand their surroundings and navigate using visual information rather than depending solely on conventional distance-mapping technology. ]]>