Vehicle Management Unit: from Model Based Design to Automatic Code Implementation & Tuning

The aim is to determine the feasibility and implementation of a vehicle management unit on a RCP or an embedded ECU, following a model-based design and code generation for implementation. With an embedded system there arises the problem of I/O interfaces as they are always adapted from system to system according to signal dynamics. The feasibility study is conducted by following the V-model scheme. Since control strategies are already well known, the thesis presents the problem of simulating the delays introduced at the I/O interfaces of an embedded system. In addition, digital filters introduce delays, which may cause the system to be unstable, mitigation of this issue is also addressed to have a well-tuned system. To understand the affects of the I/O interfaces and the signal conditioning components used to adapt the signal, a model-based scheme is used to form a base model. The base model represents the concept. In the next cycle of the development an enhanced model is produced containing the signal conditioning and conversion components, such as ADC, S/H units, to consider the delays due to I/O interfaces. Upon completion the enhanced model is validated by comparing the results with the base model. Using target hardware support package from Matlab/Simulink, ADC and DAC blocks are used as means of I/O for code generation model. The core part of enhanced model that performs the core functionality is reused to complete the code generation model and code is generated automatically. Following it, experiments are conducted to determine the exact the delays for signal traversal. The enhanced model is calibrated by adding the delay to the system model, this way the model is valid as the delays at the I/O interfaces are also taken into consideration. For tuning the digital signal at input and output, the input signal is averaged periodically and at the output the signal is smoothened. The signal is acquired at a higher frequency than the computation frequency, this way the input signal is an average value for every computation period. Furthermore, at the output the signal is linearly smoothened to have smooth signal. Also, the input signal is automatically scaled to it’s native values and units upon acquisition and then rescaled to bits after core functions are performed. All the strategies are implemented using model-based design and are compatible for automatic code generation for any target hardware, that is supported by Matlab/Simulink. In conclusion, a model is produced that is parameterized to model any target hardware with it’s specifications. However, this model will only have to be calibrated once for every target hardware. The calibration can be done by knowing the I/O delay of the target by experimenting once. For the input and output signal conditioning in code generation model, code can be generated and deployed automatically for any supported target. This makes the model reusable for any RCP or embedded ECU supported by Matlab/Simulink for code generation.