NTT Introduces Physics-Based AI Model for Improved Robot Action Prediction

NTT, Inc. has developed a new physics-based world model that allows robots to predict and evaluate potential action outcomes, such as object drops or load collapses, before execution. This technology integrates a "physics control adapter" into a video generation model, applying physical laws to calculate motion. The system significantly improved task success rates in tests, with initial applications planned for automated airport baggage handling.