Nonlinear System Identification Nano-drone Benchmark

2025-12-16

AI Summary

Modeling the dynamics of the nano-quadrotor Crazyflie 2.1 remains challenging due to real-world sensor noise, actuator nonlinearities, and inherent open-loop instability. Method: We introduce the first open-source, nonlinear system identification benchmark tailored for micro quadrotors, built upon 75k real-flight trajectories—including four aggressive maneuver types—with 4-dimensional motor commands as inputs and 13-dimensional state measurements as outputs. The benchmark supports multi-step prediction evaluation and features a novel multi-step error propagation metric for rigorous performance assessment under real hardware conditions. Contribution/Results: We release a comprehensive package—including the full dataset, experimental configurations, baseline models, and an open-source Python/Torch framework—significantly enhancing reproducibility and enabling reliable cross-method comparison. This benchmark establishes a new standard for high-fidelity modeling of agile micro aerial robots.

Technology Category

• Intelligent Robots: State Estimation
• Machine Learning: Calibration & Uncertainty Quantification
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Security and Privacy: Large-scale security measurements

Abstract

We introduce a benchmark for system identification based on 75k real-world samples from the Crazyflie 2.1 Brushless nano-quadrotor, a sub-50g aerial vehicle widely adopted in robotics research. The platform presents a challenging testbed due to its multi-input, multi-output nature, open-loop instability, and nonlinear dynamics under agile maneuvers. The dataset comprises four aggressive trajectories with synchronized 4-dimensional motor inputs and 13-dimensional output measurements. To enable fair comparison of identification methods, the benchmark includes a suite of multi-horizon prediction metrics for evaluating both one-step and multi-step error propagation. In addition to the data, we provide a detailed description of the platform and experimental setup, as well as baseline models highlighting the challenge of accurate prediction under real-world noise and actuation nonlinearities. All data, scripts, and reference implementations are released as open-source at https://github.com/idsia-robotics/nanodrone-sysid-benchmark to facilitate transparent comparison of algorithms and support research on agile, miniaturized aerial robotics.

Problem

• Develop benchmark for nonlinear system identification using nano-drone data
• Enable fair comparison of methods with multi-horizon prediction metrics
• Address challenges from real-world noise and actuation nonlinearities

Innovation

• Real-world dataset for nano-drone system identification
• Multi-horizon prediction metrics for fair algorithm comparison
• Open-source platform with baseline models and scripts