Uncertainty Quantification Framework for Autonomous System Tracking and Health Monitoring

Corbetta, Matteo; Kulkarni, Chetan; Banerjee, Portia; Robinson, Elinirina · 2021 · Crossref

DOI: 10.36001/ijphm.2021.v12i3.2936

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the critical need for systematic uncertainty quantification (UQ) in the tracking and health monitoring of autonomous and automated systems, specifically unmanned aerial vehicles (UAVs). The motivation stems from the rapid increase in low-altitude UAV operations, such as package delivery and urban air mobility, which introduces complex safety challenges. Current prognostics and health management (PHM) frameworks lack holistic approaches to account for uncertainty at the system and airspace levels. Because autonomous systems rely on predictive estimations to make self-directed decisions, ignoring uncertainty sources can lead to unsafe operations. The authors argue that predictions must be accompanied by confidence intervals, as computer models rely on simplifying assumptions that cannot replicate exact future behaviors. To address this gap, the authors propose a system-agnostic framework that maps uncertainty sources into a "predictive process structure." This methodology categorizes uncertainty into four macro-groups: Model, Method, Measure, and Input. Model uncertainty arises from abstraction (simplified physics), parameters (variable coefficients), and error (residuals between model and observation). Method uncertainty includes algorithmic limitations (e.g., convergence to local minima) and numerical errors (discretization, sampling, and coding bugs). Measure uncertainty encompasses sensor incompleteness, equipment calibration issues, and systematic errors. Input uncertainty involves initial/boundary conditions, operational requirements, and external variables like wind or temperature. The framework aims to systematically identify these sources and define how they propagate through the predictive estimation process, rather than developing new statistical techniques. The paper demonstrates the framework’s applicability through two case studies involving a small UAV. The first case study focuses on trajectory tracking, analyzing uncertainty affecting the vehicle’s adherence to a pre-defined route. The second case study addresses health diagnosis for a model-based electric powertrain, comprising a Lithium-ion battery, electronic speed controller, and brushless DC motor. These examples illustrate how the framework can be tailored to specific applications by selecting appropriate statistical techniques to handle the identified uncertainty sources. The authors note that while the case studies reflect automated systems operating on predefined instructions, the framework is equally applicable to fully autonomous systems with self-directed decision-making capabilities. The significance of this work lies in providing a structured, systematic approach to UQ for system-level PHM, which is currently missing in the literature. By explicitly linking uncertainty sources to the predictive process, the framework enables better assessment of mission success probabilities and system safety. This is particularly vital for autonomous systems, where the pool of possible operational options is theoretically infinite, leading to greater uncertainty than in automated systems. The paper concludes that integrating this framework into UAV operations will support safer integration into national airspace, minimize failure rates, and enhance the reliability of predictive diagnostics in complex, dynamic environments.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• automation surprise