Most multi-drone systems struggle with one thing: coordination under real-world constraints. A paper in Science Robotics from TU Delft proposes a model-based approach that lets multiple quadrotors jointly move and orient a cable-suspended load. BUT: without relying on learning or external sensors on the payload. Key results reported: > More than 8× acceleration of the load compared to previous state-of-the-art methods > Full obstacle avoidance while maintaining internal collision safety > Robust performance even when the load’s inertia model is uncertain > No sensors needed on the load itself Instead of data-driven models, they use a centralized trajectory optimizer that resamples past trajectories for smooth updates and stability. Each drone then executes its reference locally using onboard INDI control. It’s a reminder that not all progress in robotics comes from scaling models… some of it still comes from getting the physics and coordination right. 📍Paper: Video: —— Weekly robotics and AI insights. Subscribe free: