Aerospace and Electronic Systems - March 2019 - 8

The Adaptive Avionics Platform

CABIN AVIONICS SYSTEMS/CABIN MANAGEMENT
SYSTEMS
Current cabin systems have a centralized architecture.
Centralized controllers are connected to multiple parallel
backbone networks going through the cabin. For redundancy reasons, multiple cabin controllers exist. Cabin controllers host multiple cabin functions in parallel. DAL
levels typically range from B to D [5]. Therefore, a separation of applications must be enforced similar to IMA. Cabin
peripherals include passenger call (PC) buttons, lights, and
cabin entertainment displays. Controllers are supported by
local data concentrator devices, which translate between
local communication and a centralized backbone bus. For
most current cabin systems, the configurations of each
cabin computer are created manually or from higher order
models. This can take up to 100 days per cabin layout [14].
Cabin configurations vary per customer and undergo
frequent changes during their life time [15].

APPROACHES ADDRESSING THE AVIONICS
CONFIGURATION CHALLENGE
The common challenge for IMA and cabin systems is
configuration, including the creation of the first configuration as well as the configuration of later updates, i.e.,
changeability [16]. Current IMA and cabin architectures
require millions of configuration parameters. The manual
creation process is error prone, inefficient, or even infeasible. Solutions for the configuration challenge are
addressed in two research areas, namely model-driven
configuration generation and self-configuration.
With a domain-specific model representing the configuration data, larger datasets can be efficiently managed
due to automatic consistency and error checks [17], [18].
Moreover, less detailed architecture models can be used to
derive the detailed configuration parameters by model
transformation [19]. Model-driven configuration approaches increase the size of the configuration scenarios
that can be handled and reduce the creation time.
Self-configuring avionics determine many of the
configuration parameters on their own, depending on the
assigned application and the position in the topology. The
size and therefore the effort of external configuration files
are significantly reduced or totally abandoned. Most prominent is the Data Distribution Service (DDS), which works
with the publisher-subscriber principle. Publishers broadcast descriptions of available data with a unique ID, while
subscribers state the ID of the required data and are notified
if data are available. A communication service matches
publications and subscriptions and establishes a private
communication channel, irrespective of the physical
addresses. Moreover, DDS-based systems can change
8

communication routes during operation and are therefore
adaptive to architectural changes, such as faults. A certified
variant of DDS exists [20]. However, DDS partially moves
communication configuration into the applications, i.e., the
data descriptions and IDs. It can make the overall design
and checking of DDS-based architectures more difficult
than that of statically configured ones.
Another configuration simplification is a plug&playlike integration of peripherals into the avionics system.
For space applications, this is addressed with SPA [21]
and SPA-Z [22]. SPA enables automated peripheral discovery and self-description via the IEEE 1451 protocol
[23]. The actual identification and operation of peripherals
is, however, part of the individual system applications and
not under the control of the avionics system.
The full extent of adaptive avionics systems would be
met if the communication, peripheral integration, software
distribution, and resource sharing were organized by the
avionics system itself. One of the ultimate concepts of
adaptiveness is autonomic computing (AC) [24]. AC systems are self-configuring, self-healing, self-optimizing,
and self-protecting. AC usually addresses multiuser client-server computing environments, but the benefits for
the avionics configuration challenge are obvious [25].
There are real-time approaches for ac [26], [27]. What is
not addressed in ac is safety criticality and certification.

DIFFERENTIATION OF THE ADAPTIVE AVIONICS PLATFORM
The AAP extends existing works on self-configuring avionics by considering certification aspects and by providing
standardized failure and redundancy management.
1. All current acceptable means for certification
assume that the avionics system architecture is
quasi-static. Correctness is proven for a few welldefined system states. Countermeasures prevent any
change in the system configuration. The AAP
addresses this problem by means of differentiation
between an adaptive organization and a static standard phase.
2. Most avionics systems must exhibit redundancies
so as to increase the availability and ensure
safety. Current adaptive computing platforms
outside avionics can manage redundancies for
increasing availability, yet none of them provides
redundancy management for safety-critical systems, which first requires a robust detection of
faults and second a seamless and provably correct handover of controls to a replicate. The
AAP addresses this with a software architecture,
which can transparently handle multiple redundant execution paths. The approach is similar to
the so-called Flexible Platform [28].

IEEE A&E SYSTEMS MAGAZINE

MARCH 2019

Aerospace and Electronic Systems - March 2019

Table of Contents for the Digital Edition of Aerospace and Electronic Systems - March 2019