The feature image was bought from Istock.com and is protected by copyrights.
Aircraft are increasingly relying on software for the control of their avionics systems’ behaviour. Hence, research into more effective software engineering technologies is needed to reduce software and development complexities and to support airworthiness certifications. However, there is no detailed documentation on open source avionics software development readily available for research that can act as a benchmark to support the evaluation of proposed engineering approaches. To address this issue, we have developed the specification and design of a landing gear control software (LGCS, see Figure 1) with our own proposed methodology. In this article, we summarize the methodology and give an overview of the LGCS specification and design. We also discuss some of the issues we faced.
Requirement Specification and Design Methodology for Avionics SoftwareFigure 1. The LGCS and its operational context
Avionics software development is largely constrained by the DO-178C  (and its European equivalent ED-12C). To the best of our knowledge, no methodology in building software requirements specification and design under DO-178C has been reported in the literature. Therefore, we have defined one. We give a simplified, high-level view of this methodology in Figure 2. The process begins when a set of high-level system requirements and system safety requirements is allocated to software (SRATS). The requirements specification process covers the development (refinement and decomposition) of these SRATS into high-level software requirements (HLRs). A review of the HLRs validates these against the SRATS to check that their intent is being accurately captured by the HLRs in a way that is suitable in directing the software design process. The software design process covers the software architecture definition and the specification of low-level software requirements (LLRs). LLRs, together with the software architecture, are developed from the specified HLRs and are used to guide the source code implementation.Figure 2. General flow of the proposed methodology
The activities may be sequential but DO-178C promotes this as a way of favouring meticulous engineering to build a safe product. Nevertheless, we have organized the actions within the activities to form iterative and incremental cycles that gradually build their outputs until they yield a software requirement specification and design that is suitable in directing the following phases in the development process: coding and verification.
Developing HLRs, LLRs and Software Architecture
Avionics systems are routinely specified in natural language [2–4]. Thus, our methodology considers SRATS and HLRs to be captured using natural language. However, natural language is not a suitable form of specification to support requirements-based analyses and verification due to its inherent ambiguities . Therefore, before refining and decomposing SRATS, it is necessary to perform a manual review for ambiguities, inconsistencies and undefined conditions. This was the case with the LGCS specification. For this specification, we used as SRATS descriptions for a conventional landing gear arranged in a tricycle configuration, presented in . These descriptions were given in natural language and contained several issues. Thus, we started by correcting and improving them before developing the HLRs for the LGCS. In the interest of avoiding similar problems in the HLRs specification, we employed a form of controlled natural language. This involved specifying them in a consistent manner as statements of how the software will change a set of controllable variables (i.e. variables the software can directly affect) in response to changes in a set of monitorable variables (i.e. variables the software responds to). Table 1 shows a subset of the HLRs developed from the SRATS.
Table 1. Examples of HLRs for the LGCS
DO-178C highlights the importance of avoiding the introduction of complexities during the design process. Therefore, we incorporated principles of software engineering that help minimize complexities and promote verifiability, such as modularity and encapsulation . The result was the UML component diagram in Figure 3. We started by identifying an architectural style in line with the expected functionality of the LGCS. The closest match was the Process Control architectural style , where a system uses a set of inputs to determine a set of outputs that will produce a new state of the environment. Based on this architectural style we then defined the software components that would realize the functionality expressed in the HLRs. Each HLR was apportioned to a subset of the components, as seen in Figure 3.Figure 3. LGCS Architecture
Two parallel LLRs specifications were created: a textual specification using natural language (since they are requirements) and a model-based specification using UML state machines (since they can be regarded as the detailed software design). For the textual specification we used the same controlled natural language strategy employed with the HLRs. Table 2 shows a subset of the textual LLRs.
Table 2. Examples of LLRs for the LGCS
Figure 4 shows an excerpt of an equivalent model-based specification as a UML state machine of the LLRs in Table 2.Figure 4. Excerpt of the UML state machine for the SequenceController component
We developed the LGCS specification and design in compliance with the current avionics software regulation (DO-178C). We built them iteratively, having them verified by several domain experts. We summarize our observations throughout the process in four points:
- Quality and granularity of SRATS. We started the process with SRATS containing several inconsistencies, ambiguities and confusing wording. Hence, they had to be corrected and clarified before proceeding with the development of HLRs. Therefore, the more clear, precise and complete the SRATS are, the easier the job of the software development team becomes.
- Language used for requirements specification. We used a form of controlled natural language for both HLRs and LLRs to avoid ambiguities. LLRs were also specified using design models. Specifying HLRs using a formal language is still a challenge and is part of future work in order to support requirements-based verification.
- Granularity of LLRs. DO-178C expects LLRs to be very detailed in order to enable the implementation of source code. This may be misleading when wanting to express conditions as close as possible to the code. However, DO-178C requires a greater separation between LLRs and code than pseudocode or action languages can provide. Thus, developing the LLRs for the LGCS with an appropriate level of granularity was challenging.
- Bi-directional traceability. We used comment blocks in the design models to achieve backward traceability from software architecture and LLRs to their source HLRs (see Figures 3 and 4). However, forward traceability, e. linking HLRs to their developed LLRs, which is also required by DO-178C, was not possible with design models.
Although this is a work in progress, we have made the LGCS specification and design open and available to researchers. In doing so, we are making it possible to use it as a benchmark avionics software specification that will be used in the analysis of different engineering problems across the avionics industry and the evaluation of their proposed solutions.
For more information on this research, please refer to the following conference paper:
Andrés Paz and Ghizlane El Boussaidi. Building a Software Requirements Specification and Design for an Avionics System: An Experience Report. In Proceedings of SAC 2018: Symposium on Applied Computing, Pau, France, April 9–13, 2018 (SAC 2018). DOI: 10.1145/3167132.3167268
Andrés Paz is a PhD student in the Department of Software Engineering and IT at ÉTS. He is working on a DO-178C-compliant model-driven approach to support the development and certification of safety-critical avionics software.
Program : Software Engineering
Research laboratories : LASI – Computer System Architecture Research Laboratory
Ghizlane El Boussaidi
Ghizlane El Boussaidi is a professor in the Department of Software Engineering and IT at ÉTS. She specializes in model-based software engineering, and software architecture and design.
Research laboratories : LASI – Computer System Architecture Research Laboratory