Preliminary design of future reconfigurable IMA platforms

Pierre Bieber, Eric Noulard, Claire Pagetti, Thierry Planche, Francois VialardSigbed rev 2009 · DOI

TL;DR

This paper proposes a preliminary design for future Integrated Modular Avionics (IMA) platforms that can dynamically reconfigure themselves to improve aircraft operational reliability and reduce maintenance costs. The authors investigate architecture principles and reconfiguration services, focusing on how to manage application migration when hardware components fail.

What Problem Does This Paper Address?

Modern aircraft rely heavily on Integrated Modular Avionics (IMA) platforms to consolidate numerous avionics functions onto fewer, shared computing modules. While this consolidation reduces costs and weight, it also means that a single hardware failure can impact multiple critical applications. The current generation of IMA (IMA-1G) has limited fault tolerance. This paper addresses the need for future IMA platforms (IMA-2G) to incorporate reconfiguration capabilities. The goal is to automatically adapt the system's configuration by moving applications from a faulty module to a spare one, thereby enhancing operational reliability (minimizing flight delays or cancellations due to hardware issues) while maintaining safety and minimizing the impact on certification processes.

Methods

The paper proposes a reconfiguration framework built around a "Reconfiguration Supervisor." This supervisor is responsible for detecting failures, determining if a reconfiguration is necessary (i.e., if the aircraft becomes "NOGO"), and then selecting and executing a new, safe configuration.

Key aspects of their approach include:

  • Reconfiguration Scenarios: They define scenarios based on granularity (module vs. partition level), location (local vs. distant spares), and timing (on-ground vs. in-flight).
  • Reconfiguration Services: These include functions for data loading, monitoring and fault detection, and power supply management, all orchestrated by the supervisor.
  • Configuration Representation: A configuration is defined as a function mapping applications to computing elements (module and partition).
  • Reconfiguration Policies: These are off-line rules that guide the selection of the next configuration, considering factors like application priority and spare module ordering.
  • Graph of Reconfigurations: They model possible reconfiguration sequences as directed acyclic graphs, allowing for analysis of reachable states and transitions under different policies.
  • Resource and Real-time Constraints: Configurations must satisfy resource availability (CPU, memory) and segregation requirements. Transitions must be safe and bounded in duration, with a target of under 15 minutes for on-ground reconfigurations.
  • Continuity of Service: The reconfiguration process must be transparent to other running applications, ensuring uninterrupted service.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Reconfiguration as a Core Service: Unlike traditional fault tolerance where reconfiguration is a recovery mechanism, this paper positions it as a fundamental capability for future IMA platforms, aiming to proactively improve reliability.
  • The "NOGO" Trigger: A critical insight is that reconfiguration is only triggered when a failure has a direct impact on operational readiness (making the aircraft "NOGO"), rather than any minor fault.
  • Policy-Driven Reconfiguration: The choice of reconfiguration policy significantly impacts the complexity and number of possible configurations and transitions, influencing the feasibility of certification.
  • Certification Challenges: The sheer number of potential configurations in a reconfigurable IMA system poses a significant challenge for traditional certification methods, suggesting a need for model-based approaches.
  • Supervisor Redundancy: The authors note that the supervisor itself can be a single point of failure, highlighting the need for a redundant supervisor to ensure high operational reliability.

Key Results

The paper presents a preliminary design and analysis rather than quantitative experimental results. Key findings are conceptual:

  • They define the formal representation of configurations and analyze the number of possible configurations and transitions for module-level reconfiguration under two distinct policies (spare reconfiguration disallowed vs. allowed).
  • They establish that the policy allowing spare reconfiguration leads to a more complex, non-tree graph of reconfigurations with potentially more transitions.
  • They identify that the supervisor's availability is a critical factor in the overall reliability improvement.

Limitations and Future Work

  • Preliminary Design: The paper presents a preliminary design, and a real demonstrator is being developed within the SCARLETT project.
  • Certification: The authors acknowledge that certifying a system with a large number of configurations is a major challenge, suggesting model-based safety assessment as a potential solution.
  • Supervisor Reliability: The single point of failure of the supervisor is identified as a limitation, suggesting the need for a redundant supervisor.
  • Future Work: The authors plan to implement on-ground module and partition-level reconfigurations, detail the analysis of the proposed solution concerning operational reliability, safety, and certification, and explore in-flight reconfiguration and reconfiguration for safety in the longer term. They also suggest studying other scenarios and leveraging the ASAAC standard for distributed implementations.

Summary updated

BibTeX
@article{bieber2009preliminary,
  author = {Pierre Bieber and Eric Noulard and Claire Pagetti and Thierry Planche and Francois Vialard},
  title = {{Preliminary design of future reconfigurable IMA platforms}},
  journal = {Sigbed rev},
  year = {2009},
  doi = {10.1145/1851340.1851349},
}
Figure