Purpose

This paper aims to overcome the main obstacle to compare the merits of the different control strategies for fixed-wing unmanned aerial vehicles (UAVs) to assess autopilot performances. Up to now, the published studies of control strategies have been carried out over disperse models, thus being complicated, if not impossible, to compare the merits of each proposal. The authors present a worked benchmark for autopilots studies, consisting of generalized models obtained by merging UAVs’ parameters gathered from selected literature (journals) with other parameters directly obtained by the authors to include some relevant UAVs whose models are not provided in the literature. To obtain them it has been used a dedicated software (from U.S. Air Force).

Design/methodology/approach

The proposed models have been constructed by averaging both the main aircraft defining parameters (model derivatives) and pole-zero locations of longitudinal transfer functions. The suitability of the used methodologies has been checked from their capability to fit the short period and the phugoid modes. Previous analytical model arrangement has been required to match a uniform set of parameters, as the inner state variables are neither the same along the different published models nor between the additional models the authors have here contributed. Besides, moving models between the space state representation and transfer function is not just a simple averaging process, as neither the parameters nor the model orders are the same in the different published works. So, the junction of the models to a common set of parameters requires some residual’s computation and transient responses assessment (even Fourier analysis has been included to preserve the dominance of the phugoid) to keep the main properties of the models. The least mean squares technique has been used to have better fittings between SISO model parameters with state–space ones.

Findings

Both the SISO (Laplace) and state-space models for the longitudinal transfer function of an “averaged” fixed-wing UAV are proposed.

Research limitations/implications

More complicated situations, such as strong wind conditions, need another kind of models, usually based on finite element method simulation. These particular models apply fluid dynamics to study aerostructural aircraft aspects, such as flutter and other aerolastic aspects, the behavior under icing conditions or other distributed parameter problems. Even some models aim to control other aspects than the autopilot, such as the trajectory prediction. However, these models are not the most suitable for the basic UAV autopilot design (early design), so they are outside the objective of this paper. Obviously, the here-considered UAVs are not all the existing ones, but the number is large enough to consider the result as a reliable and realistic representation. The presented study may be seen as a stepping stone, allowing to include other UAVs in future works.

Practical implications

The proposed models can be used as benchmarks, or as a previous step to produce improved benchmarks, in order to have a common and realistic scenario the compare the benefits of the different control actions in UAV autopilots continuously presented in the published research.

Originality/value

A work with the scope of the presented one, merging model parameters from literature with other (often referred in papers and websites) whose parameters have been obtained by the authors has been never published.

中文翻译：

---

用于自动驾驶控制器设计的无人机广义纵向模型

目的

本文旨在克服比较固定翼无人机 (UAV) 不同控制策略的优点以评估自动驾驶性能的主要障碍。到目前为止，已发表的控制策略研究是在分散模型上进行的，因此比较每个建议的优点是很复杂的，如果不是不可能的话。作者提出了一个自动驾驶研究的工作基准，包括通过合并从选定文献（期刊）收集的无人机参数与作者直接获得的其他参数获得的广义模型，以包括一些相关的无人机，其模型未在文献中提供。为了获得它们，它使用了专用软件（来自美国空军）。

设计/方法/方法

所提出的模型是通过对主要飞机定义参数（模型导数）和纵向传递函数的零极点位置进行平均来构建的。所用方法的适用性已经从它们适应短期和长周期模式的能力进行了检查。以前的分析模型安排需要匹配一组统一的参数，因为内部状态变量在不同的已发布模型中或在作者在这里贡献的其他模型之间都不相同。此外，在空间状态表示和传递函数之间移动模型不仅仅是一个简单的平均过程，因为在不同的已发表作品中参数和模型阶数都不相同。所以，将模型连接到一组通用参数需要进行一些残差计算和瞬态响应评估（甚至包括傅立叶分析以保持 phugoid 的优势）以保持模型的主要特性。最小均方技术已被用于在 SISO 模型参数与状态空间参数之间进行更好的拟合。

发现

提出了用于“平均”固定翼无人机纵向传递函数的 SISO（拉普拉斯）模型和状态空间模型。

研究限制/影响

更复杂的情况，如强风条件，需要另一种模型，通常基于有限元法模拟。这些特定模型应用流体动力学来研究航空结构飞机方面，例如颤振和其他气动弹性方面、结冰条件下的行为或其他分布参数问题。甚至一些模型旨在控制自动驾驶仪以外的其他方面，例如轨迹预测。然而，这些模型并不是最适合基本无人机自动驾驶设计（早期设计）的，因此它们不在本文的目标范围内。显然，这里考虑的无人机并不是所有现有的，但数量足够大，可以将结果视为可靠和现实的表示。所提出的研究可能被视为垫脚石，

实际影响

所提出的模型可以用作基准，或作为产生改进基准的前一步，以便有一个通用和现实的场景来比较已发表的研究中不断提出的无人机自动驾驶仪中不同控制动作的好处。

原创性/价值

一项具有所提出的范围的工作，将文献中的模型参数与作者获得的其他参数（通常在论文和网站中提到）合并，但从未发表过。