VTOL technology is moving beyond the air taxi race, with civilian infrastructure and autonomous military aircraft emerging as two key areas of development. bit.ly/4yeof1l

Researchers at Nvidia have developed SONIC, a foundation model for humanoid robot movement that acts as a universal "motor system" — translating high-level commands from VR teleoperation, video-based motion capture, and vision-language-action models into smooth, coordinated whole-body robot movements, all without retraining between tasks. The controller was trained on more than 100 million frames of human motion data, allowing it to learn broad movement patterns rather than individual skills. In tests, SONIC successfully reproduced both previously learned and unseen movements using a single model, rather than the separate control policies typically required for each task. "Today's systems are trained with separate controllers for different skills, making them difficult to scale," said Yuke Zhu, director and distinguished research scientist at Nvidia. "We wanted to explore whether a single learned controller could serve as a foundation for many forms of whole-body motion." The team also demonstrated remote robot teleoperation and the conversion of written instructions into physical robot actions. Nvidia plans to integrate SONIC with its Isaac GR00T platform, combining high-level AI reasoning with general-purpose whole-body control. Target applications include manufacturing, warehousing, and logistics. The research is published in Science Robotics. 📷 Luo et al., Science Robotics / NVIDIA