Embedded and Real-time Systems

Embedded and Real-time Systems

Ahmet Onat
  • Embedded Systems

    Although greater than 95% of all computers in use today are embedded computers, people are generally unaware of their presence. Main lines of research today in embedded systems are in validation of performance, enhancing processor architecture to meet real-time requirements, research in real-time operating systems, real-time networked systems, and enhancing security. Research is being conducted on most of these areas at Sabanci University. Most embedded systems are in control of a physical plant, hence they must complete their decisions within a time limit. In some applications such as auomotive brakes, these time limits must never be missed, making the validation and testing of real-time systems an important research area at SU.

    The performance of embedded systems differ from general purpose computer systems where the time required to make decisions based on current inputs (latency) is more important than number of operations per second (throughput). Embedded hardware design focuses on such areas as multi-threaded processors and decreasing latency while keeping the operation of the processor deterministic.

    Real-time operating systems schedule many tasks on the processor while making sure that none miss their deadlines. Researchers work to discover new algorithms which meet these goals, aided by hardware designers who produce processors according to their specifications.

    Modern manufacturing systems are becoming extremely complex. A single but complex computer system to control them is not very practical because the amount of computation required would be too great. Use of several network controlled smaller computers offers an efficient solution. To meet the real-time requirements, however, all of the computers and the network must have time bounds to complete calculations and the data transfers. This area of networked control systems is new and thriving; seen by most researchers as the future of embedded systems. Such systems will also infiltrate our life under such names as ambient intelligence.

    Sensitive data should be protected from being monitored and modified by third parties. This can be done using crypto algorithms. However, since the real-time requirements put a bound on the amount of computation we can perform before decisions must be made, sufficiently secure crypto algorithms that do not load the processor must be discovered. Overall, the research on embedded systems is paving the way to a society which uses resources efficiently. It is one of the key research areas that is both challenging and one that has direct applications, and results which are urgently needed.

  • Networked Contol Systems

    In networked control systems, the components of the control loop reside on different computer nodes, and are connected by a communication network. Many modern control and automation systems are networked. Some examples are cars (with around 50 computer nodes around the typical car), automation and instrumentation systems (production machinery, railway systems, ships etc.) utilities (such as water works of a city) and many similar systems. Many consumer devices are becoming networked, with networked control being their natural extension.

    The fundamental difficulty in networked control is the unreliability of the communication network where information may be delayed or completely lost. This makes it difficult to guarantee the stability of the plant. We have proposed "Model Based Predictive Networked Control System" (MBPNCS), which overcomes this difficulty by making use of a local model of the plant within the controller node, and a protocol which obviates the need for reverse flow of acknowledge information through the network. The work progresses in theoretical framework, simulations and exoeriments on real-time computers.