# Learning Cross-Coupled and Regime-Dependent Dynamics for Aerial Manipulation

- Source: arXiv cs.RO
- Published: 2026-10-19T04:00:00.000Z
- RoboHot: https://robohot.news/en/items/m100kecpk9y1onfrxxrewi7v3
- Original: https://arxiv.org/abs/2605.14805

## Summary

A study proposes a structured encoder-decoder framework in which a nonlinear history encoder captures the temporal and cross-coupled residual dynamics of an aerial manipulation platform and its robotic arm, while a linear latent-space decoder supports rapid adaptation at deployment time. The framework's linear-in-parameters structure enables closed-form Bayesian updates and integration with real-time model predictive control (MPC). Experiments on a real aerial manipulation platform showed improved residual prediction accuracy, rapid adaptation after dynamics changes, and reduced MPC tracking error under changing payload conditions.