This thesis focuses on model-based testing of embedded software systems. A particular model-based testing approach called model-based statistical testing is chosen for this purpose. This approach allows for the automated generation, execution and evaluation of test cases based on a test model. It is additionally possible to estimate the reliability of the tested system based on test results. These properties are desirable in embedded software system testing. An important property of an embedded software system is that its inputs and outputs are continuous and discrete signals and that it is essential that these signals need to fulfill real-time constraints. The signals are measured from and applied to a concurrent environment additionally. Model-based statistical testing as described in literature is not applicable to systems with these properties i.e. it is not applicable for testing embedded software systems. This is why this thesis focuses on improving model-based statistical testing. This is achieved by the introduction of a new test model called usage net.