A methodology for estimating the region of attraction for autonomous nonlinear systems is developed. The methodology is based on a proof that the region of attraction can be estimated accurately by the zero sublevel set of an implicit function which is the viscosity solution of a time-dependent Hamilton–Jacobi equation. The methodology starts with a given initial domain and yields a sequence of region of attraction estimates by tracking the evolution of the implicit function. The resulting sequence is contained in and converges to the exact region of attraction. While alternative iterative methods for estimating the region of attraction have been proposed, the methodology proposed in this paper can compute the region of attraction to achieve any desired accuracy in a dimensionally independent and efficient way. An implementation of the proposed methodology has been developed in the Matlab environment. The correctness and efficiency of the methodology are verified through a few examples.
The flight envelope plays an important role in flight safety. The concept of posing the flight envelope as a region of attraction is explored further, and it is investigated whether the stable manifold for the region of attraction computation is an efficient method for determining envelope. The stable manifold describes the flight dynamic envelope of an aircraft in an explicit representation, which means that the computation needs to be done only on the envelope, not the entire state space. In this paper, the stable manifold is computed by using a fast method which reduces the computation to solving a system of partial differential equation. Then, the stable manifold grows in the way of advancing front mesh generation framework. The stable manifold is then applied to the envelope determination of a nonlinear F-16 model. The result is compared to the results obtained with the level set method, demonstrating that the stable manifold provides a feasible and accurate result to the dynamic envelope. The proposed method is then used to investigate the effect of actuator failure on the flight safety. The proposed method can also be used as a safety assessing tool during the design phase of an aircraft.
A flight safety warning method based on reachability analysis and fuzzy inference is proposed against aircraft icing. A nonlinear model of iced aircraft with uncertainty is proposed based on the existing research results on the effects of icing and the uncertainty in icing detection. To deal with the uncertainty caused by icing, reachability analysis is used to estimate the safe flight envelope of iced aircraft. On this basis, fuzzy inference is employed for flight safety warning which can be used to enhance the pilot’s situational awareness in icing encounters. Simulations of the GTM (Generic Transport Model) aircraft show that, the proposed method has the potential to further increase the flight crew awareness about the risk of losing control in flight under icing.
With the rapid development of AI technology in the low-altitude field, the predictive uncertainty of ML models, particularly object-detection models, has become a key bottleneck restricting their airworthiness and safe deployment. A typical manifestation of this is that when facing test samples in open-world scenarios, object-detection models exhibit overconfidence in erroneous predictions. To address this issue, this paper proposes an anomaly-scoring algorithm for out-of-distribution (OOD) evaluation based on augmented deep network features, named PCA-HBOS. By integrating the high-dimensional semantic features extracted by deep networks, the algorithm can score sample distributions, thereby enabling the identification of both in-distribution and out-of-distribution samples. Through comparisons with mainstream OOD algorithms, the superiority of the PCA-HBOS in low-altitude scenarios is validated. Experimental results on three multi-sensor UAV detection datasets—MSDD, AOT, and AOD-4—demonstrate that, except for Faster R-CNN and YOLOv2, the PCA-HBOS achieves favorable OOD discrimination performance across multiple object-detection models. On false-negative samples, it attains optimal AUC and FPR95 values of 0.995 and 0.291, respectively, while on false-positive samples, the optimal values reach 0.960 and 0.354. Furthermore, the experiments further explore the relationship between model generalization and OOD algorithm performance. The proposed algorithm can provide methodological support for model performance testing and optimization, contributing to the development of intelligent UAV systems towards safer and more trustworthy directions.