DroneDiffusion: Robust Quadrotor Dynamics Learning with Diffusion Models

2024-09-17

arXiv.org

AI Summary

To address the control fragility of quadrotor systems arising from modeling errors and external disturbances, this paper proposes a dynamics learning framework based on conditional diffusion models, framing dynamics modeling as a conditional sequence generation task—the first application of diffusion models to uncertainty modeling in UAV dynamics. Departing from conventional Gaussian or deterministic assumptions, our approach explicitly captures multimodal, time-varying uncertainties and their temporal evolution. Integrated with adaptive controller design and rigorous stability analysis, it effectively suppresses error propagation. Simulation and real-world flight experiments demonstrate substantial improvements in trajectory tracking accuracy and system robustness under challenging conditions—including unknown trajectory tracking, variable payload, high-speed maneuvers, and strong wind disturbances—while exhibiting strong generalization capability.

Technology Category

• Intelligent Robots: State Estimation
• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Diffusion Models for Vision

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

An inherent fragility of quadrotor systems stems from model inaccuracies and external disturbances. These factors hinder performance and compromise the stability of the system, making precise control challenging. Existing model-based approaches either make deterministic assumptions, utilize Gaussian-based representations of uncertainty, or rely on nominal models, all of which often fall short in capturing the complex, multimodal nature of real-world dynamics. This work introduces DroneDiffusion, a novel framework that leverages conditional diffusion models to learn quadrotor dynamics, formulated as a sequence generation task. DroneDiffusion achieves superior generalization to unseen, complex scenarios by capturing the temporal nature of uncertainties and mitigating error propagation. We integrate the learned dynamics with an adaptive controller for trajectory tracking with stability guarantees. Extensive experiments in both simulation and real-world flights demonstrate the robustness of the framework across a range of scenarios, including unfamiliar flight paths and varying payloads, velocities, and wind disturbances.

Problem

• Addresses quadrotor fragility from model inaccuracies and disturbances
• Overcomes limitations of deterministic or Gaussian uncertainty assumptions
• Enables robust dynamics learning for complex real-world scenarios

Innovation

• Uses conditional diffusion models for dynamics learning
• Integrates learned dynamics with adaptive controller
• Captures temporal uncertainties for robust generalization