• AVR
• Engine Control
• Real Time
• Scheduling

The modern Engine Control Unit (or ECU) has become a complex and powerful electronic device, being able to manage all fundamental aspects of passenger cars.
A typical ECU is characterized by computational activities that are periodic or triggered by the rotation of the crankshaft, thus generating a workload variable with the engine speed. To prevent overloads, a common practice with angular tasks is having a certain number of operational modes implemented, with different computational cost, that are activated at different speed intervals: for that reason they are also referred as Adaptive Variable Rate (AVR) tasks. Understanding how different choices of switching speeds and computational loads influence the behaviour of the engine is fundamental to obtain the best performances of the system.
This thesis presents a mathematical model of a Diesel engine, created in Simulink environment, that is used to test the impact of Real-Time scheduling on a set of performance indices (thermodynamic efficiency, pollutant emission, noise). First, the schedulability problem for AVR task sets is presented, with the theoretical basis and state of the art. The mathematical model is then illustrated in detail: the dynamic of the engine is modelled using a simplified set of physical equations taken from specialized literature, while the controller is inspired by the real ECU using a mix of maps and robust control laws. The scheduler is integrated in the Simulink model using the T-Res framework (a RTSim Simulink interface), which has been modified to manage also the AVR tasks in its features. Finally, a series of tests are made, varying the switching speeds and control strategies, to show how performance indices are influenced. An experimental validation of the model is also presented to demonstrate the applicability of these results.