A Decentralized Algorithm for Radio Resource Management in Heterogeneous Wireless Networks with Dynamic Number of Mobiles Julio Rojas-Mora * , Habib B.A. Sidi * , Rachid El-Azouzi * , Yezekael Hayel * * LIA, Université d’Avignon Avignon, France Email: [julio.rojas, habib.sidi, rachid.elazouzi, yezekael.hayel]@univ-avignon.fr Abstract Overlapping coverage of multiple base stations provides a multiple degree of freedom problem for mobiles which have to choose between different wireless technologies. This paper presents a framework to achieve an efficient dynamic and decentralized radio resource management in heterogeneous wireless network scenarios. Each mobile has to decide to which technology it wants to be connected to. In such problems of user-network association, we consider non-cooperative games and we show the existence and uniqueness of Nash equilibrium under some assumptions on the utility functions. We design a fully distributed algorithm that can be employed for converging the system to a Nash equilibrium. The novelties of the proposed algorithm are that, under dynamic environment (mobiles may leave or join the system), this algorithm is still robust and provides fast convergence to Nash equilibrium. Moreover, if there is a new arrival, each mobile already connected to a base station always stays connected which avoids repeated vertical handovers. Also, the new arrival runs the proposed algorithm in order to find the best base station to be connected to. For the departures, the algorithm reaches a Nash equilibrium using as few vertical handovers as possible, with as few mobiles as possible; increasing the speed of convergence. Finally, we apply our algorithm for different overlapping situations between heterogeneous wireless networks. I. I NTRODUCTION In the world of wireless networks, the emergence of new technologies induce that mobiles will have the possibility to have access to different technologies at the same location. Indeed, with the growing offers for 3G+ access everywhere from telephony providers, more and more users equipped with laptops, netbooks or new generation smartphones can have simultaneously access to WiFi and 3G. Moreover, the emergence of wide area wireless networks based on WiMAX or LTE will complicate the allocation problem for the mobiles. Over the last years there has been a growing interest on how users should be assigned to exploit those resources more efficiently. Therefore, there have been many proposals based on joint Radio Resource Management (RRM) to efficiently share the scarse network resource in order to take the advantage of all the heterogeneous resources available. But resource management for QoS provisioning in heterogeneous wireless network is a complex task because of heterogeneity of policies and mechanisms used for QoS provisioning in different wireless networks, along-with highly probabilistic behavior of network traffic. Moreover, the presence of the multiple radio access technology provides roaming capability in different wireless networks through vertical handovers. Those handover operations may cause significant degradation to QoS provisions. Mechanisms for optimizing and controlling mobiles in an area where there is coexistence of wireless technologies is studied in [5] with cognitive radios WiFi cells inside a WiMAX cell. Another work is described in [2] with the technology 802.11(e) and a central controller. Those studies propose a centralized approach. Some papers propose a joint radio resource (JRRM) mechanism to achieve an efficient usage of a joint pool of resources. For example, [7] and [8] propose a framework for a JRRM based on fuzzy neural methodology and reinforcement learning algorithms. Reinforcement learning techniques have been first applied in a wireless network for studying optimal power control LIA Technical Report 01-10. This work has been partially supported by the French national project ANR WiNEM, ANR-06-TCOM-05.