On Distributed Control-by-Wire Systems for Critical Applications

Doctoral thesis, 2005

The implementation of closed loop control systems is an important field for the exploitation of embedded real-time computer systems. One can find a particularly interesting class of closed loop control applications within vehicle dynamics control. For example, within automotive electronics engineering we refer to drive-by-wire control systems as an automotive class of applications where there are no mechanical, pneumatic or hydraulic physical connections between the steering wheel, pedals, wheels and engine control. We commonly refer to such solutions as control-by-wire applications. Control objects in ground vehicles such as passenger cars and railway cars are in most cases physically distributed. A distributed computer system solution therefore becomes an intuitive choice for the computer control. When inherent redundancy is then present due to the distribution of the control system, the objective becomes to take full advantage from it. This avoids introducing excessive redundancy for fault-tolerance, thus giving a cost-efficient implementation. In this thesis, we present a fault-tolerant computer architecture suitable for critical distributed control. Based on the architecture, we further propose a computer-based brake-by-wire control system where the design alternatives are heavily restricted by cost-efficiency requirements. The thesis gives a novel approach to the implementation of cost efficient dependable electronic systems. The basic ideas of taking advantage of application inherennt redundancy are adopted in a non-complex hardware device. Simulation of the device has shown suitability for intended purposes. The resulting overhead of complexity introduced by this device has been analysed; cost and performance has been estimated. The demands put upon the real-time communication system are significant for emerging distributed control applications. The paradigm of distributed information processing adopted within distributed control has an important impact on the underlying communication system. We start with identifying requirements on the communication network from a control system point of view. Then, we outline a research platform for evaluation of communication protocols. Recent real-time communication protocols support both event-triggered and time-triggered communication. Thus, we may anticipate hybrid scheduling methods for future engineering needs. Therefore, we discuss valid criteria for practical usefulness of contemporary real-time communication protocols. We identify methods for timing analysis and scheduling of communication between nodes in critical systems.