• performance
• real-time control
• simulation

Engine Control Unit (ECU) is characterized by computational activities that are activated with different mechanisms. Some engine control tasks are activated at a fixed rate (Periodic task), whereas other tasks are triggered at specific crankshaft angles (Angular task). A potential problem with such type of activities is that, for high speeds, the computing system that executes the control algorithm can suffer overload conditions. A common practice to prevent such problems is to implement angular tasks with a set of executing mode such that they automatically decrease their computational requirements for increasing speeds. For this reason they are also referred as Adaptive Variable Rate (AVR) task. The ECU of a commercial Diesel car can be seen as a Real-Time system where all its control tasks must be completed within a deadline. However, in reality, exceeding a deadline would not cause a failure of the system but instead can lead to a loss of performance introducing errors or delays in the actuation. For this reason the analysis of how timing issues could influence the behavior of the engine is fundamental to improve the performance of the system. The first part of this thesis has been focused on the model of the combustion process of a four-cylinder Diesel engine, capturing the interaction among each cylinder dynamic and handling multiple injection strategies, in order to obtain a precise behavior of the combustion process. Moreover, the dynamics of a certain number of performance parameters have been developed, with particular emphasis on the pollutant emissions. These models have been integrated in the framework presented in [20], improving also the control part and adding other physical components, like the Exhaust Gas Recirculation and the gearbox. The global framework has been used for analyzing how timing effects in the scheduling part can affect the performance parameters. The multiple injection model provides the first justification for an adaptive task execution behavior and a study of how different injection strategies can affect the performance is carried out. The plant is controlled with a mix of static maps, inspired by the real engine applications whereas two antagonistic PID controllers are applied for the Variable Geometry Turbocharger (VGT) control. Both the physical system and the control parts are implemented in Simulink environment while the scheduler is integrated in the simulation using the TRES framework, which is a Simulink interface of the open source scheduler simulator RTSim, which has been developed at the Retis Lab, entirely in C++. A series of simulation results are illustrated with particular focus on how different strategies of injection can impact the performance of the system. In the end different deadline misses' patterns are implemented in order to analyze how timing issues can influence the injection and finally the performance indexes.