Embedded real-time systems are those systems which have many components interacting with each other and with the environment. In this thesis, we introduce a methodology to generate test cases for embedded real-time systems based on test purposes expressed as Message Sequence Charts (MSCs), and timed input output automata as specification model. The approach consists of six main operations: (1) the parsing of the specification and test purposes; (2) the selection of the transition paths to be considered for test cases generation; (3) the construction of the synchronous product of the test purposes and the TIOA specifications; (4) the construction of a partial product for the system under test; (5) the sampling of TIOAs; and (6) the generation of test cases. Within these six operations, the sampling operation can be executed in different levels. Each level gives rise to a new method for test cases generation for embedded real-time systems. We implemented and studied three of these methods by comparing them in terms of the number of states generated and the number of test cases devised.