Photo credit Saab A modern aircraft hosts lots of advanced software and electronics (avionics) in the form of sensors that gather information, compute units where the information is processed, actuators that control the aircraft, and equipment that presents information to the pilot. During operation, this information is updated repeatedly, giving rise to a complex flow of data between different units, and thus placing requirements on when different activities are allowed to be executed.
Avionic systems in modern aircraft must be flexible and scalable
The avionic industry has moved towards an architecture called Integrated Modular Avionics (IMA), where applications share processing and communication resources. This enables savings in weight and energy consumption, and most importantly, it brings flexibility and scalability. This is critical in future avionic systems due to the very rapid development of intelligent systems. For example, if it would be too expensive to update a software application, or too difficult to install additional computing resources or sensors, the whole system would quickly become obsolete.
Photo credit Saab
Central for an IMA-system is to assure that different applications do not disturb each other. This can be accomplished through a spatial partitioning (allocation of hardware) along with a temporal partitioning (scheduling) of the system. In a distributed integrated avionic system, there is a platform that provides memory and runtime to the software applications and also handles all communication between different software systems and integrated external sensors. Here, the industry is moving towards modular architectures with well defined interfaces and with a clear separation between the abstract model that applications are developed for, and the specific hardware on which a system runs. Two examples are FACE and Pyramid. A key technology is Time Sensitive Networking (TSN), a collection of standards for communication over Ethernet, that for example allows scheduled traffic and best effort traffic side-by-side, on the same network. Another key technology is virtualized hardware, in the form of virtual machines, which enables flexible partitioning of the physical computing resources.
New mathematical models and algorithms are required to fully utilize the systems
These flexible techniques and standards have many benefits, but in order to effectively utilize the resources and the flexibility of the system, a number of complex resource allocation and scheduling decisions need to be handled. This comes from that all activities, including traffic in an TSN network, must be scheduled. Further, processors must be partitioned in virtual machines in a suitable way and the activities distributed among them. Both resource allocation and scheduling can be handled by solving NP-hard discrete optimization problems, but the scale and the structure of the problems make solving them extremely difficult.
Using standard software for solving this kind of optimization problems is not a realistic option, and the lack of suitable methods for resource allocation and scheduling may inhibit continued technological development. Thus, to fully utilize the inherent potential in these avionic systems, new mathematical models and algorithms are required that are tailored for the specific scheduling and resource allocation problems that come with the systems.
The project is a collaboration between Saab Aeronautics and Linköping University and is partially founded by the National Aeronautics Research Programme (NFFP).