NTT unveils AI system to predict cargo behavior for robot handling

NTT announced a physical AI system that enables robots to assess how cargo will react to their actions, helping prevent damage, according to a report by Mobile World Live link. The company says the proprietary physical control adapter evaluates the corporeal properties of objects and a deep-learning image generation model visualises likely reactions, adding a prediction layer to existing robot control systems.

Current robot handling approaches, NTT notes, suffer from inaccuracies and rely heavily on large training datasets, limiting performance in unfamiliar scenarios. In internal tests, a small robot simulating cargo stacking achieved a near 20% improvement in success rates using the new system.

NTT plans to apply the technology first to airport luggage handling and lists logistics, warehousing, manufacturing, construction and retail as additional sectors. A partnership with robotics software provider Mujin is slated for future demonstrations, aiming to embed the NTT World Model into broader physical AI deployments.

Our read: The system addresses a known gap in robot-assisted material handling by reducing dependence on exhaustive training data, potentially lowering operational losses from cargo damage across multiple B2B verticals.