Neural-Augmented Incremental Nonlinear Dynamic Inversion for Quadrotors with Payload Adaptation

2025-03-12

AI Summary

For quadrotor systems with suspended payloads, residual forces are difficult to model accurately, conventional controllers neglect their effects, and incremental nonlinear dynamic inversion (INDI) suffers from high sensitivity to noisy sensor measurements—particularly in derivative estimation. To address these challenges, this paper proposes a neural-enhanced INDI control framework. Its core innovations include: (i) replacing the noise-sensitive numerical differentiation in INDI with a lightweight neural network for robust derivative estimation; (ii) establishing a tightly coupled integration architecture that fuses neural prediction with INDI; and (iii) the first systematic adaptation of this framework to the dynamics modeling and real-time control of quadrotors with underactuated suspended loads. Experimental results demonstrate that, without additional hardware, the method significantly improves residual force estimation accuracy. Compared to standard INDI, it reduces trajectory tracking error by 32%, accelerates disturbance rejection response by 41%, and enhances payload swing suppression by 57%.

Technology Category

• Intelligent Robots: State Estimation
• Search and Optimization: Mixed Discrete/Continuous Search
• Humans and AI: Planning and Decision Support for Human-Machine Teams

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

The increasing complexity of multirotor applications has led to the need of more accurate flight controllers that can reliably predict all forces acting on the robot. Traditional flight controllers model a large part of the forces but do not take so called residual forces into account. A reason for this is that accurately computing the residual forces can be computationally expensive. Incremental Nonlinear Dynamic Inversion (INDI) is a method that computes the difference between different sensor measurements in order to estimate these residual forces. The main issue with INDI is it's reliance on special sensor measurements which can be very noisy. Recent work has also shown that residual forces can be predicted using learning-based methods. In this work, we demonstrate that a learning algorithm can predict a smoother version of INDI outputs without requiring additional sensor measurements. In addition, we introduce a new method that combines learning based predictions with INDI. We also adapt the two approaches to work on quadrotors carrying a slung-type payload. The results show that using a neural network to predict residual forces can outperform INDI while using the combination of neural network and INDI can yield even better results than each method individually.

Problem

• Accurate flight control for quadrotors with payload adaptation.
• Predicting residual forces using learning-based methods.
• Combining neural networks with INDI for improved performance.

Innovation

• Neural network predicts smoother INDI outputs
• Combines learning-based predictions with INDI
• Adapts methods for quadrotors with payloads