Receding-Horizon Control via Drifting Models

1. 1Learn a trajectory generator through distribution matching using an offline dataset of trajectories
2. 2Use the learned generator to plan trajectories in real-time, adapting to changing system dynamics
3. 3Employ receding-horizon control to optimize the planned trajectory and minimize a desired cost function
4. 4Refine the model and control strategy through iterative updates and feedback from the system

This research benefits AI engineers and ML researchers working on control systems and trajectory optimization, as it provides a novel approach to handling uncertain dynamics

💡 The approach enables trajectory optimization in settings where system dynamics are unknown and simulation is not possible

💡 Receding-Horizon Control via Drifting Models optimizes trajectory planning in unknown dynamics