We investigate, in this thesis, the use of Hierarchical Modulation (HM), a physical layer technique that enables to exploit multiuser diversity, for resource allocation in OFDMA-based systems with and without use of relaying, so as to improve the system capacity. HM allows the sharing of the resources, namely subcarriers and power, between users of different radio conditions by sending an additional stream to a user with good radio conditions on a subcarrier that was initially allocated to carry an original stream to a user with lower radio conditions. And this, without affecting the original user's rate nor the total amount of power assigned to the shared subcarrier. In the literature, most of the works that consider the use of HM focus solely on the physical layer performance, notably in terms of the bit error rate. And this for a static user scenario,i.e., with a fixed number of users in the system, each with an infinite service duration. This configuration however does not reflect the real system behavior where the number of users is dynamic, i.e., the users come to the system at random time epochs and leave it after a finite duration, corresponding to the completion of their services. The study of the system at the flow-level, as opposed to the packet level, for a dynamic user configuration, enables us to investigate the realistic relationship between capacity and demand and to quantify several system-level performance metrics, such as mean transfer times and blocking rates, which are meaningful both to the user and the network operator/provider.