A Sliced Learning Framework for Online Disturbance Identification in Quadrotor SO(3) Attitude Control

Fri, 06 Feb 2026 00:00:00 +0000

This work introduces a dimension-decomposed geometric learning framework called Sliced Learning for disturbance identification in quadrotor geometric attitude control. Instead of conventional learning-from-states, this framework adopts a learning-from-error strategy by using the Lie-algebraic error representation as the input feature, enabling axis-wise space decomposition (“slicing”) while preserving the SO(3) structure. This is highly consistent with the geometric mechanism of cognitive control observed in neuroscience, where neural systems organize adaptive representations within structured subspaces to enable cognitive flexibility and efficiency. Based on this framework, we develop a lightweight and structurally interpretable Sliced Adaptive-Neuro Mapping (SANM) module. The high-dimensional mapping for online identification is axially “sliced” into multiple low-dimensional submappings ( “slices”), implemented by shallow neural networks and adaptive laws. These neural networks and adaptive laws are updated online via Lyapunov-based adaptation within their respective shared subspaces. To enhance interpretability, we prove exponential convergence despite time-varying disturbances and inertia uncertainties. To our knowledge, Sliced Learning is among the first frameworks to demonstrate lightweight online neural adaptation at 400 Hz on resource-constrained microcontroller units (MCUs), such as STM32, with real-world experimental validation.

Introduction Video:

For testbed experiment video:

Online Identification using Adaptive Laws and Neural Networks for Multi-Quadrotor Centralized Transportation System

Thu, 13 Nov 2025 00:00:00 +0000

This paper presents an online adaptive–neuro identification framework to enhance the robustness of centralized multi-quadrotor cable-suspended payload transportation systems. The proposed method addresses the challenges arising from highly coupled nonlinear dynamics, unknown payload parameters, and time-varying disturbances by introducing a dimension-decomposed learning strategy.

The core idea is to decompose the high-dimensional payload error space into multiple low-dimensional subspaces, referred to as “slices.” Within each slice, shallow neural networks and adaptive laws operate in parallel to identify model uncertainties and disturbance dynamics online. This structure, termed Sliced Adaptive-Neuro Mapping (SANM), enables efficient real-time learning without requiring persistent excitation or offline training, while preserving theoretical transparency.

The SANM module is integrated as a feedforward compensator into a payload-centric geometric control architecture. By learning directly from Lie-algebra–based error representations, the proposed framework avoids explicit modeling of complex cable–quadrotor couplings and improves disturbance rejection for both translational and rotational payload dynamics. Stability of the overall closed-loop system is rigorously analyzed using Lyapunov theory, and uniform ultimate boundedness of tracking and estimation errors is established under unknown payload parameters and time-varying disturbances.

Numerical simulations demonstrate that the proposed method significantly enhances tracking performance and robustness compared with conventional centralized control approaches. The results indicate that dimension-decomposed adaptive–neuro learning provides an effective, lightweight, and theoretically grounded solution for robust aerial cooperative transportation.

Article-Journal | Tianhua Gao

A Sliced Learning Framework for Online Disturbance Identification in Quadrotor SO(3) Attitude Control

Online Identification using Adaptive Laws and Neural Networks for Multi-Quadrotor Centralized Transportation System