DROPS

Document

DOI: 10.4230/OASIcs.DX.2024.16

Quantifying the Sim-To-Real Gap in UAV Disturbance Rejection

Authors: Austin Coursey, Marcos Quinones-Grueiro, and Gautam Biswas

Published in: OASIcs, Volume 125, 35th International Conference on Principles of Diagnosis and Resilient Systems (DX 2024)

Abstract

Due to the safety risks and training sample inefficiency, it is often preferred to develop controllers in simulation. However, minor differences between the simulation and the real world can cause a significant sim-to-real gap. This gap can reduce the effectiveness of the developed controller. In this paper, we examine a case study of transferring an octorotor reinforcement learning controller from simulation to the real world. First, we quantify the effectiveness of the real-world transfer by examining safety metrics. We find that although there is a noticeable (around 100%) increase in deviation in real flights, this deviation may not be considered unsafe, as it will be within > 2m safety corridors. Then, we estimate the densities of the measurement distributions and compare the Jensen-Shannon divergences of simulated and real measurements. From this, we show that the vehicle’s orientation is significantly different between simulated and real flights. We attribute this to a different flight mode in real flights where the vehicle turns to face the next waypoint. We also find that the reinforcement learning controller actions appear to correctly counteract disturbance forces. Then, we analyze the errors of a measurement autoencoder and state transition model neural network applied to real data. We find that these models further reinforce the difference between the simulated and real attitude control, showing the errors directly on the flight paths. Finally, we discuss important lessons learned in the sim-to-real transfer of our controller.

Cite as

Austin Coursey, Marcos Quinones-Grueiro, and Gautam Biswas. Quantifying the Sim-To-Real Gap in UAV Disturbance Rejection. In 35th International Conference on Principles of Diagnosis and Resilient Systems (DX 2024). Open Access Series in Informatics (OASIcs), Volume 125, pp. 16:1-16:18, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2024)

Copy BibTex To Clipboard

@InProceedings{coursey_et_al:OASIcs.DX.2024.16,
  author =	{Coursey, Austin and Quinones-Grueiro, Marcos and Biswas, Gautam},
  title =	{{Quantifying the Sim-To-Real Gap in UAV Disturbance Rejection}},
  booktitle =	{35th International Conference on Principles of Diagnosis and Resilient Systems (DX 2024)},
  pages =	{16:1--16:18},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-356-0},
  ISSN =	{2190-6807},
  year =	{2024},
  volume =	{125},
  editor =	{Pill, Ingo and Natan, Avraham and Wotawa, Franz},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.DX.2024.16},
  URN =		{urn:nbn:de:0030-drops-221087},
  doi =		{10.4230/OASIcs.DX.2024.16},
  annote =	{Keywords: sim-to-real, disturbance rejection, unmanned aerial vehicles}
}

Document

Short Paper

DOI: 10.4230/OASIcs.DX.2024.21

Data-Driven RUL Prediction Using Performance Metrics (Short Paper)

Authors: Abel Diaz-Gonzalez, Austin Coursey, Marcos Quinones-Grueiro, Chetan S. Kulkarni, and Gautam Biswas

Published in: OASIcs, Volume 125, 35th International Conference on Principles of Diagnosis and Resilient Systems (DX 2024)

Abstract

Prognostics is the scientific study of component and system degradation with use, and the prediction of when failure may occur. In this work, we propose a new data-driven method for predicting a system’s remaining useful life (RUL) without needing an accurate system model or expert knowledge. Instead, we use system operational data to estimate how the system’s performance metrics change with time. Although this is a purely data-driven approach, the method’s design is inspired by model-based techniques. First, we frame a novel Multitask Machine Learning architecture to simultaneously learn the general pattern of performance degradation and the individual trajectories from run-to-failure performance trajectory data. We apply this method to the set of performance metrics that determine the system’s end-of-life (EOL), building a performance trajectory library of the system operation under different operational conditions. We leverage the performance metric library as prior belief and develop a Bayesian deep learning approach to update the performance measures over time and predict the system EOL. We evaluate our method on two datasets of the N-CMAPSS benchmark, achieving satisfactory results in terms of overall performance and uncertainty estimation accuracy. Overall, our approach illustrates a generalized deep learning architecture that can more effectively predict the system RUL for a collection of identical systems.

Cite as

Abel Diaz-Gonzalez, Austin Coursey, Marcos Quinones-Grueiro, Chetan S. Kulkarni, and Gautam Biswas. Data-Driven RUL Prediction Using Performance Metrics (Short Paper). In 35th International Conference on Principles of Diagnosis and Resilient Systems (DX 2024). Open Access Series in Informatics (OASIcs), Volume 125, pp. 21:1-21:15, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2024)

Copy BibTex To Clipboard

@InProceedings{diazgonzalez_et_al:OASIcs.DX.2024.21,
  author =	{Diaz-Gonzalez, Abel and Coursey, Austin and Quinones-Grueiro, Marcos and Kulkarni, Chetan S. and Biswas, Gautam},
  title =	{{Data-Driven RUL Prediction Using Performance Metrics}},
  booktitle =	{35th International Conference on Principles of Diagnosis and Resilient Systems (DX 2024)},
  pages =	{21:1--21:15},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-356-0},
  ISSN =	{2190-6807},
  year =	{2024},
  volume =	{125},
  editor =	{Pill, Ingo and Natan, Avraham and Wotawa, Franz},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.DX.2024.21},
  URN =		{urn:nbn:de:0030-drops-221135},
  doi =		{10.4230/OASIcs.DX.2024.21},
  annote =	{Keywords: remaining useful life, data-driven methods, machine learning, performance metric, multitask machine learning, Monte Carlo}
}

Search Results

Documents authored by Coursey, Austin

Quantifying the Sim-To-Real Gap in UAV Disturbance Rejection

Abstract

Cite as

Data-Driven RUL Prediction Using Performance Metrics (Short Paper)

Abstract

Cite as

Thanks for your feedback!

Could not send message