The development of third generation mobile cellular networks, and more recently of fourth generation mobile communication systems, has been promoted by the growth of mobile voice and data in parallel with the advances in radio technology for handheld devices, including smart phones, tablet computers and other platforms of audio-visual media. Enabled by Moore's law and the resulting increase in available hardware performance, mobile equipment running over multiple radio standards has been devised. Mobile Internet has been made possible due to advancements in mobile and wireless communications and the packet switch nature of the Internet which allows users to always be connected at any time without a-priori reservation of (expensive) network resources. At the same time, the primary goal for the evolution of the Internet was to enable peer-to-peer communication among computers of any network. This goal has been achieved by means of the Internet Protocol (IP) which when combined with a set of transport protocols transformed the Internet to a network of networks. This protocol suite, known as TCP/IP, provided the foundations for the technical progress during the last two decades and created new types of services and new possibilities of conducting communications and media activities. Prominent applications for conveying information, like the World Wide Web, allowed computer users to locate and view multimedia content (e.g. text, audio, graphics, animation, and video) resulting in an increasing difficulty to distinguish between the IT, telecommunications and media markets. In fact, the same service can today be implemented by different technologies, depending on the medium or the channel of communication used for conveying it to the recipient in a manner in which the recipient obtains the service. This phenomenon is referred to as convergence in terms of both technology and business. Technology convergence refers to a trend in which some technologies having distinct functionalities evolve to a technology accommodating all distinct functionalities simultaneously. Convergence of technologies enables and in some cases even forces business enterprises formerly active in a particular sector to extend their product portfolio and businesses in other related sectors. In telecommunications all major sectors of businesses and technologies including services, devices and networks have been subjects to convergence. The convergence of each one of the technologies is defined in terms of the other two.
Device convergence refers to the combination of different types of devices originally designed for different services and networks into a unified device able to communicate through a variety of networks and support any multimedia service. The development of mobile phones to devices supporting functionality besides telephony, such as camera, video, and Internet and to multi-mode terminals enabling multiple access methods simultaneously, such as CDMA, and Wi-Fi, is an illustrative example of multi-access multi-service device convergence.
Services convergence refers to the merging of different types of services originally associated with specific devices and supplied by different media into a communication service supplied by any device and medium. Telephony, initially implemented on circuit-switched networks, and TV, originally conveyed on broadcasting networks, can be now delivered as packets to any IP-enabled device of any access technology by means of voice over IP (VoIP) and IPTV respectively.
Network convergence refers to the consolidation of different networks - also with respect to their characteristics and functionality - originally developed for different services and devices into an integrated network capable of carrying any service for a variety of devices. For the mobile systems there is a trend towards a coupling and cooperation among cellular networks and their complementing wireless access systems such as wireless local networks (WLAN), wireless personal networks (WPAN), mobile ad-hoc networks (MANET), wireless sensor networks (WSN), digital broadcasting networks, and the Internet.
Based on these trends of convergence, and with the IP as one of the key enabling technologies it is envisioned that future broadband mobile networks will consist of an IP-based packet communications infrastructure offering converged services. The convergence to the IP-based networking world has recently been accelerated by advancements in the provision of access and core bandwidth for both mobile and fixed networks. Especially within the radio access networks, the initial integration steps of heterogeneous radio technologies into one multi-radio access environment have been taken both by standardisation bodies and commercial operators. Next-generation wireless systems aim at combining the wide coverage of the cellular networks with other higher capacity radio networks to better meet the increasing volumes of data traffic generated by an increasing number of mobile users and higher-bandwidth consuming services. These objectives involve both the development of new radio technologies, which aim at further improving spectral efficiency, but also the cooperation among existing or new radio access technologies. By means of cooperation the heterogeneity of the underlying radio access networks should be transparent to both mobile users and services. Further integration opens the potential to access any network providing an "Always Best Experience" for end users, in an "always best connection" fashion . It also opens the possibility for mobile devices to be served via different service access points (i.e. relays, infrastructure access points) possibly belonging to different radio access networks (RANs) implemented by different radio access technologies (RATs) and operated by different operators. A RAT is defined by the type of radio technology and its corresponding wireless interface between any two communicating nodes. The co-existence of different radio technologies e.g., 3GPP GSM, UMTS and LTE, IEEE 802.11a/b/g/n and 802.16 at the same location provides a multi-radio heterogeneous radio environment where radio diversity gains can be sought. Diversity in conjunction with mechanisms for a flexible use and efficient management of radio access resources, (e.g., selection of a "best" type of access), serve both users (e.g. low cost versus high performance) and providers (e.g. load sharing). Furthermore, maintaining connection and service continuity and mobility for users moving between accesses requires support for handover between different RATs, support for cooperation among the different RANs and support for rapid establishment of roaming agreements (dynamic roaming) and efficient announcing strategies (of both user needs and provider offers).
Ambient Networks Approach
Deploying a patchwork of IP-based solutions is not sufficient to fully benefit from the potential of convergence. It requires the deployment of a dynamic, co-operative and business-aware consistent network control layer. This was the rationale behind the Ambient Networks approach. The Ambient Networks (AN) project - developed a particular networking vision for Beyond-3G and 4G based on the cooperation of heterogeneous networks characterised by different radio access technologies, radio access networks and user/operator domains. A basic mechanism of the AN concept is, hence, the dynamic and instant "composition" of networks which is best described as a state of inter-network agreements between ANs that enable cooperation. The automation of the network composition and cooperation allows the usage of resources without the need for pre-configuration or detailed offline negotiation between network operators. The vision of the AN approach was to device a horizontal control space consisting of functional elements that provide communication and content services between networks and nodes based on network composition.
With the above network vision as a requirements basis, the multi-radio access (MRA) architecture , has been designed with the ability to control and utilize more than one radio access resources for the transmission of user data. The architecture is novel in its provisioning of functionalities and mechanisms to support the AN vision of a dynamic environment with a multitude of different wireless devices, network operators and business actors that can form instant inter-network agreements with each other. The AN MRA proposes a framework and an architecture that allows joint radio resource management and utilisation across different RATs at various levels of integration. This is mainly achieved by means of Multi-Radio Access Selection (MRAS) and provided by the Multi Radio Resource Management (MRRM) , and the Generic Link Layer (GLL) , components. These are built on previous research, such as - and , respectively, generalised and extended with functionality to address the requirements posed by the dynamic ambient networking. MRRM is responsible for joint management of radio resources and load sharing between the different RANs and GLL provides a toolbox of configurable link layer functions that allows cooperation between different RATs.
In particular, MRRM performs Layer-3 management of data flows, dealing with events with "long-term" dynamics that are comparable with the duration of the data flow. Main Layer-3 functionalities include broadcast of system information, cell selection, paging, establishment, maintenance and release of connection, measurement reporting and control, mobility management. In the context of MRA, a key function of MRRM is the selection and configuration of multiple radio access technologies (RATs) for a mobile terminal's data flow. RAT selection may be changed several times during the life-time of a data flow. GLL performs Layer-2 Medium Access Control and Radio Link Control, dealing with events with "short-term" dynamics that are comparable with PHY-layer transmission times. Main Layer-2 functionalities include dynamic scheduling and ARQ. In the context of MRA, a key function of GLL includes fast access selection at a packet-level as a means of a dynamic scheduling of mobile terminals' data flows across the RATs selected by the MRRM. This mechanism exploits the benefits from diversity of multiple radio accesses and is referred as multi-radio transmission diversity (MRTD).
The advent of IP in computer networks and the introduction of packet switch paradigm in the mobile cellular networks underpinned the development of the "all IP networks" vision. The "all IP" vision introduced the Internet design principles in mobile networks, provided the means for a common network protocol to transport user data, and enabled the design and the deployment of converged networks. At the same time the existence of multiple radio access technologies ranging from medium to short-range, like WLAN and Bluetooth, to mobile cellular, like, UMTS, and the deployment of multi-mode multi-standard terminals underpinned the development of the "always best connected" concept . The idea was simply to have the multi-radio standard terminals connected to the best radio interface at any time. The best connection is the one that can support the QoS demands of a user's traffic and can be provided to the lowest cost.
The concept of "always best connected" targeted the marriage of two key features that radio technology by late 90s could not offer in a single RAT, namely, high bit rate transmission and terminal mobility. High bit rates could be provided in relatively smaller coverage areas by means of e.g., IEEE 802.11 a/b/g/n -. The small coverage area is a result of the design of the radio access protocols employed. Also the absence of mobility between cells is due to the missing radio resource management functionality which is deemed expensive and unnecessary for technologies operating in unlicensed spectrum. On the other hand, mobile cellular networks, e.g., UMTS, are dimensioned for estimated peak demands and cell layouts are usually driven by the full coverage need. The overlap among neighbouring cells is limited to what is necessary for the mobility management. The wide coverage areas of the cellular networks are rather served by moderate bit rates as compared to 802.11 but can increase to a RAT-specific maximum rate in areas where base station density is highly increased  . The advent of the "all IP" concept was long deemed as the enabler for a seamless interworking between radio access networks where the "always best connected" can be realised. However, operating at the IP network level was insufficient to handle the control of the converged networks and to support the always best connected paradigm. At first, utilisation and mobility between multiple radio access technologies requires the unification of three basic network mechanisms: (i) mobility, (ii) security and (iii) Quality of Services (QoS).
Mobility refers to the ability of users to freely move within and between RATs without interruption of the user's connections to other peers or applications across the Internet. The term mobile here implies that all connections are automatically maintained despite changes in the user's point of attachment to the network which are caused due to the user's movement. Maintaining or seamlessly re-establishing a connection requires a context transfer mechanism which facilitates the exchange of information about the connection between the two points of attachment (or access points). In the mobile cellular networks the technical process by which a user device changes its point of attachment is called handover or handoff. Within RATs standardised by 3GPP  organisation, handover occurs between cells which are the points of attachment in the mobile cellular networks.
Security refers to an integrated Authentication, Authorisation and Accounting mechanism used to grant access to and control and keep track of utilisation of network resources by the users. Authentication is the process of determining the identity of the user, and ensuring that the user is a valid user. Authorisation is the process of granting (or denying) access to users, and permitting users to a certain utilisation of resources. Finally, accounting is the process of monitoring users' activities while accessing network resources, including time resources used, service time, amount of data transferred during a connection. Accounting data is used for e.g., billing, auditing and cost allocation. Security services often require a server that is dedicated to providing the three services.
Quality-of-Services (QoS) refers to the performance specification of a connection or a service. Typically, the performance is indicated quantitatively in terms of service-related Key Performance Indicators (KPI). Typically, different layers define their own KPIs e.g., Bit Rate (BR), Bit Error Rate (BER) and Block Error Rate (BLER), are widely used in radio link layer whilst at network layer it is more common with throughput, packet delay, packet error rate and packet loss.
The unification of mobility, security and QoS mechanisms has been studied and proposed in numerous projects within the 5th EU IST Framework programme. However, the unification of mobility, security and QoS alone could not guarantee efficient utilisation of local radio resources among the radio access technologies. To this end multi-radio coordination control became the research topic of numerous research projects, including Ambient Networks within the 6th EU IST Framework programme. The main focus was to extend the functionality of Radio Resource Management to facilitate and manage resources from multiple radio access technologies. One of the key objectives was to increase spectral efficiency and to better meet users QoS demands and enable Mobile Internet anytime and anywhere. With the development of the 3GPP Long Term Evolution (LTE)  that made Mobile Internet feasible by a single RAT the focus of the 7th EU IST Framework programme projects has shifted. Despite this shift multi-radio access is still considered within R&D projects, standards and products. The main results of these projects and activities are shortly summarised in the subsequent sections.
R&D projects - Prior state-of-the-art
All projects preceding this work developed multi-access architecture solutions by adding necessary functionality on the IP network protocol which was used as the common denominator. This approach was mainly adopted by leading IST projects within the EU FP5 framework programme. The IST project MIND  and its predecessor BRAIN  have designed an all-IP mobile access network, with QoS and IP-based micro mobility support within a single radio access. For the inter-domain/inter-access mobility, Mobile IP is employed. Mobile IP (or MobileIP)  is an open standard communication protocol defined by the IETF (Internet Engineering Task Force)  that enables mobile device users to keep the same IP address while roaming between IP networks. Mobile IP works by permitting a mobile node to have two IP addresses: a home address IP (maintained and stored by the domain's Home agent) and a "care-of" address IP (assigned and advertised by the visiting domain's foreign agent). The home agent is responsible to encapsulate and redirect all received packets for a mobile device to the device's new care of address. While Mobile IP can be used to support interdomain mobility, it cannot cope with fast handovers between access points within a domain, especially in networks where handovers are frequently occurring e.g., small pico cellular networks. As opposed to Mobile IP inter-domain interacting, IP micro mobility implies protocols that are designed to support intra-access mobility of users and to enhance QoS by effectively reducing delay and packet loss during handoff. In terms of layering, a generic service interface termed 'IP to Wireless' (IP2W) carried by a convergence layer between IP and Radio Link have been devised to support seamless hand-over between heterogeneous access technologies and to handle application QoS demands. An overview of the proposed architecture is provided in  and .
A similar approach has been also adopted by the Wireless Internet Networks (WINE) . In dealing with the performance enhancement of Internet protocols over WLANs, the core of the WINE proposal is the wireless adaptation layer (WAL) that provides a uniform interface to IP, while adapting to the higher layers' requirements and observed channel conditions. In its simplest form seamless network access can be realised by a network management server implementing an AAA system, a home agent for IPv6 mobility management, a QoS broker and an IP paging agent. This architecture that utilises IP for mobility management, QoS provisioning and AAA as a convergence layer has been proposed by the MobyDick project . The principle to leave the radio access systems unaltered and to provide (via the backbone) mechanisms and solutions for cooperation between access systems has been followed by the majority of the projects in the EU FP5 including the DRiVE project  and its successor the OverDRIVE . The DRIVE project introduces an overlay design for the network architecture consisting of a part of individual radio access systems and an access independent backbone part providing inter-system mobility and a security server that provides AAA services using the DIAMETER  protocol. In addition the backbone facilitates multi-access coordination based on Dynamic Spectrum Allocation (DSA) negotiations between different Radio Access Networks (RANs) (and possibly different operators). The WINEGLASS project  deepens the integration by using IP technology in the interior of the UMTS mobile network. Session mobility and authentication is no longer duplicated at both IP and UMTS. As a result, the GPRS nodes of the core network are not necessary and replaced by border routers which are directly connected to Radio Network Controllers (RNCs).
In summary the architectures for the systems beyond 3G - using IP to combine 3G, wireless LANs and other access technologies - have designed an all-IP mobile access network, with new middleware components and protocols to align QoS and support seamless hand-over between heterogeneous access technologies. A comparison analysis of the most representative architectures is given in . A radio abstraction layer with some similarities to GLL but narrower in scope has been defined in these projects. The GLL extends these concepts, e.g. concerning "lower-level" link layer functions and multi-radio multi-hop support. Furthermore GLL extends the IP mobility concepts by including options for access path switching at Layer 2. These projects also investigated IP mobility schemes for inter-system handovers. MRRM further extends the IP mobility with e.g. radio resource management considerations in inter-RAT mobility.
R&D projects - Contemporaneous and contemporary state-of-the-art
WINNER, E2R, EVEREST and its successor AROMA are projects that like AN conducted research on various aspects of multi-access under the umbrella of the EU IST 6th Frame Programme. In this section we give a short summary of the objectives and approaches of these focusing on issues common to MRA.
The main objectives of WINNER , are-among others-to "define a single ubiquitous radio access system concept, scalable and adaptable to different short range and wide area scenarios" and, more relevant from a GLL-MRTD point of view, "define radio level co-operation mechanisms between different Radio Access Networks (RANs)". However, according to  no RA cooperation mechanisms such as MRTD are investigated. WINNER focuses more on mobility management (handover and location based RRM), and common RRM functions like admission control, scheduling/load control and QoS based management. The GLL concept is viewed as an approach the WINNER radio interface should principally be compatible to.
The key objective of the E2R project , is to devise, develop and trial architectural design of reconfigurable devices and supporting system functions in the context of heterogeneous mobile radio systems . There are some similarities between concepts of E2R and the concepts in Ambient Networks. The Joint Radio Resource Management (JRRM) concept of E2R, which is applied for heterogeneous access networks and multi-radio terminals, covers different management layers and service types. JRRM operates on short term but as well on medium and long term timescales. Accordingly medium and long term input information (e.g. coverage/ availability of RAT, current load in each RAT, expected traffic characteristics) and as well as short term radio channel variations are considered by JRRM. Two of main building blocks of the JRRM architecture are the Joint Session Admission Control (JOSAC), and the Joint Resource Scheduler (JOSCH) which functionality resemble GLL MRTD in switched and parallel mode respectively.
One of the overall goals of EVEREST  is to develop Common RRM (CRRM) algorithms mainly between access technologies deployed in Universal Terrestrial Radio Access Network (UTRAN) and GSM/EDGE Radio Access Network (GERAN) where both tight and very tight coupling are considered. Other access technologies and networks like WLANs and future radio technologies are considered as well. In  the access selection approach is based on cost functions defined by service and network factors and operator preferences. These cost functions are applied at large timescales (e.g. per service, cost, day time) covering mainly the network layer (RRM) up to service layer (charging).
The objectives of the AROMA project  converge with the Ambient Networks-like MRTD goals. In AROMA shorter time scales associated with radio channel fluctuations and tighter coupling between RATs are investigated to exploit the trunking gains and thereby improve performance. To address the heterogeneity of the wireless RATs the access selection decision is further based on a "fittingness" factor reflecting on the degree of adequacy of a given RAT to a given service requested by a given user. Results from simulations show considerable gains in terms of QoS. 
Finally, the End-to-End Efficiency (E3) project  aimed at the introduction of cognitive wireless systems in the Beyond 3G (B3G). The proposed B3G cognitive system framework accommodated solutions for optimising the use of spectrum and radio resources by means of cognitive network and collaborative decision-making methods between network elements and terminals. Towards this end, a collaborative cognitive radio resource management including dynamic spectrum management (DSM), joint radio resource management (JRRM) for inter-technology cooperation and self-organisation has been developed. .
Feasibility studies and technical requirements for seamless multi-radio access systems have been produced in all major global standardisation bodies IETF, 3GPP and IEEE 802, Despite their legacy in fixed IP networks, cellular system and WLAN, respectively, they all focus towards an integrated network consisting of several accesses that are connected to a common IP core. In such an integrated network the utilization of the accesses across the different networks and the provision of the services and applications have to be done seamlessly.
The IEEE working group (WG) 802.21  was formed to specify media independent handover (MIH) and interoperability. The first meeting was in March 2004 and consequently the work was running almost in parallel with the AN project. The 802.21 standard  describes an architecture that supports transparent services between heterogeneous access networks for mobile terminals by means of mobility management protocols. The reference architecture describes MIH functions between legacy lower layers (L1 and L2) and upper layers (L3 and above). This "Layer 2,5" functionality resembles the idea and scope of GLL. As in GLL, link events can indicate changes in state and transmission behaviour of the L2 data links, including resource management. The MIH specification defines command services, trigger events and information services that are provided by the L2 link for optimal handover performance. Apart from reducing handover latency, the specification of such an abstraction layer can also assist in network discovery and selection, but a description of a general multi-radio resource management entity is not explicitly given.
IP mobility schemes and extensions have been developed and standardised by several Working Groups of IETF . Among the most relevant, the Context Transfer, Handoff Candidate Discovery, and Dormant Mode Host Alerting Working Group (Seamoby WG)  considered issues that were related to multi-radio access, while the Mobile IPv4 Working Group  and Mobile IPv6 Working Group , specified protocols to provide fast handover solutions for IPv4 and IPv6 respectively. In particular the Seamoby developed mechanisms assumed that a set of candidates has already been chosen and that handover can be initiated to all of the suitable candidates. In addition to mobility, the PILC WG  developed guidelines for the efficient design of link layer protocols. The guidelines target good Internet services and in particular services that utilize the Transmission Control Protocol (TCP). These guidelines are of particular importance when dynamic access selection is enabled and changes of the access systems occur frequently. Recently, the Multi-path TCP (MPTCP) WG  addresses at TCP level the simultaneous utilization of many different paths between peers in a similar manner as it has been proposed by means of MRTD and GLL. The protocol provides the components necessary to establish and use multiple TCP flows across potentially disjoint paths.
The 3GPP group  defined the All-IP Network (AIPN) concept , which must support both GERAN and UTRAN based access networks as well as other networks like WLANs and WPANs. The scope is to identify requirements associated with the evolution of 3GPP Systems towards AIPN and find requirements for the reuse of legacy infrastructure. The multi-access interworking should be handled by a centralised mobility manager. To facilitate inter system handover over several access network technologies 3GPP has introduced in its Evolved Packet Core (EPC) definition - the Access Network Discovery and Selection Function (ANDSF) component . EPC, which is destined for the 4G technologies aims at integrating 3GPP as well as non-3GPP access network technologies. To this end ANDSF, similarly to IEEE's MIH, aims at facilitating handovers, in particular the inter system handovers. Another 3GPP work with relevance to MRA, is the network sharing concept. In , architectures to be supported by network sharing are specified. A network sharing architecture allows for different operators to connect to a shared UMTS network. The operators are not limited only to share the radio network elements, but may also share the core network and the radio resources themselves.
The above solutions have been tailor-made for combinations of two specific RATs at a time, and having classical operator-user relations in mind. The AIPN provides a means for a general multi-radio access solution embracing new and other legacy RATs and allowing for new ways of cooperation between operators and users. In particular, the 3G/WLAN integration developed by 3GPP can be considered as one simplistic case of integration. Similarly, the UTRAN/GERAN/LTE cooperation can be seen as a simple sub-case of performing MRRM. In general, within 3GPP, a tighter integration between access systems and networks is viable among 3GPP compatible technologies which can be further enhanced by the introduction of multi-standard base stations . Another enhancement of the common resource management within 3GPP can be achieved by means of dynamic spectrum allocation. 3GPP LTE-Advanced facilitates carrier aggregation that allows for the aggregation of the system capacity ,.
However, the related work listed above differs in several ways to the proposed MRRM and GLL functions of the MRA and the implementation of MRTD. To our best knowledge, at the time of their conception no prior solution addressed a joint radio resource management in conjunction with a tight integration at the link layer level.
Summary of Contributions
This thesis is a compilation of seven conference papers each one included in Part II. Contributions may be divided in two major groups: (i) MRA architecture design and system specification for MRTD and GLL, and (ii) performance evaluation of MRTD. The first group of contributions are described in chapters 2 and 3 and correspond to papers 1-3 while the second group on feasibility studies, corresponding to papers 4-7, are summarised in chapter 5. To the set of concepts and ideas presented in this work, the major conception contributions of the author of this thesis are the overall GLL layered architecture, as depicted in chapter 2, and the definition, and performance evaluation of the Multi-radio Transmission Diversity concept, as given in chapters 3 and 4. It has to be acknowledged that MRTD has been jointly conceived, and to equal extent originated and developed by the author of this thesis and Dr. Hamid Reza Karimi. The description of MRTD in chapter 3 partially follows the original common contribution provided to the AN project . The reminder of this section highlights the main contributions of the author of this thesis.
MRA architecture and MRTD
Scope: The general scope of the contributions is as follows.
The development of multi-radio access selection mechanisms for optimised resource utilisation in terms of multi-radio transmission diversity;
the design of a multi-radio access architecture involving the selection mechanisms and functionality;
the specification of the functionality implementing multi-radio transmission diversity as part of a more generic link layer functionality.
Research method: The foundation of the research approach has been based on the analysis of a large set of requirements and objectives, followed by a synthesis of the emerged concepts, and system features, all structured into a consistent functional architecture supporting multi-radio access utilisation.
 J. Sachs, L. Muñoz, R. Agüero, J. Choque, G. Koudouridis, R. Karimi, L. Jorguseski, J. Gebert, F. Meago, and F. Berggren, "Future Wireless Communication based on Multi-Radio Access," in Proc. Wireless World Research Forum WWRF11, 10-11 June 2004.
Summary: This paper introduces multi-radio resource management and generic link layer which are the main concepts and mechanisms for the utilisation of multiple radio accesses in Ambient Networks. The different types of requirements related to system architecture, system performance and end-user control necessitate a modular functionality which allows combinations of interfaces and functions to form different levels of cooperation and collaboration between different RATs. Each level of cooperation is defined in terms of interactions among and between MRRM and GLL instances and their associated groups of functions. The grouping of the functions and the open interfaces are illustrated by means of a logical node architecture and a functional protocol layer architecture showing protocol terminations.
Contribution by author: The author of this thesis contributed in this paper as follows: (i) authored and contributed to the definition of the GLL concept section, (ii) was the originator of the underlying modular MRA architecture as depicted in Figure 1 of the paper and the author of the corresponding Functional Protocol Layer Architecture section, (iii) contributed to the discussion about the collaboration across different radio technologies, and (iv) reviewed the paper and contributed to the clarity in the description of the ideas.
 K. Dimou, R. Agüero, M. Bortnik, R. Karimi, G.P. Koudouridis, S. Kaminski, H. Lederer, J. Sachs, "Generic link layer: a solution for multi-radio transmission diversity in communication networks beyond 3G," in Proc. IEEE Semi-annual Vehicular Technology Conference, VTC-2005-Fall, vol. 3, pp. 1672- 1676, 25-28 September 2005.
Summary: This paper proposes the system architecture and the GLL functions required for the realization of multi-radio transmission diversity. The basic functions that MRTD consists of are: (i) radio access selection, (ii) performance monitoring and (iii) flow and error control. These three functions, their interactions and their location within a multi-radio access network are further elaborated and described. The technical discussions include the implementation of both MRTD at IP and at MAC levels and their applicability in the context of cooperation between UMTS/HSPA and WLAN.
Contribution by author: The author of this thesis contributed in this paper as follows: (i) authored and contributed with the specification and the design of the MRTD functionality section, (ii) co-authored the MRTD realisation section and contributed to the cooperation between 3G and WLAN with the definition of access selection, flow control and multi-radio ARQ functions at both general and IP-level, (iii) contributed to the conclusions and discussions section, and (iv) reviewed the paper and suggested improvements in the presentation of the ideas.
 G.P. Koudouridis, P. Karlsson, J. Lundsjö, A. Bria, M. Berg, L. Jorguseski, F. Meago, R. Agüero, J. Sachs, R. Karimi, "Multi-Radio Access in Ambient Networks," in Proc. EVEREST-Workshop2005, 16 November 2005.
Summary: In this paper the functions and the interactions between M-RRM and GLL for the radio access selection in the Multi-radio Access (MRA) architecture are described along with a series of feasibility studies evaluating their performance. The feasibility studies are defined based on scenario assumptions that take into consideration (i) the impact of different levels of multi-RAT coupling to the radio access selection performance gains ranging from fast access selection to slow access selection, (ii) the cooperation among multiple operators ranging from competitive to fully cooperative scenarios, (iii) the impact of aged information and signalling delays as in the case of non-collocated RATs, and (iv) other critical factors as the availability of limited backhaul level of cooperation among RATs. The results from a series of evaluation studies are presented emphasizing the scenarios where MRA approach shows significant gains in the overall system performance. The results from the feasibility studies are summarised and potential solutions that are envisaged to improve further the MRA performance are outlined.
Contribution by author: The author of this thesis contributed in this paper as follows: (i) had editor responsibility and authored the paper based on input from the co-authors, (ii) contributed to the analysis of the multi-radio access selection algorithms, and (iii) contributed to and summarised the performance evaluation studies on MRTD..
MRTD feasibility studies
Scope: The general scope of the contributions is the evaluation of the multi-radio transmission diversity performance by means of simulations based on a theoretical setting and on practically simplified though indicative deployment scenarios.
Research method: Mechanisms for multi-radio access selection have been explored by means of feasibility studies. More specifically, in this thesis performance evaluation studies on the MRTD functionality have been performed by means of analysis and simulations. The evaluation was performed in different scenarios and settings where the number of available radio accesses, the network load and the network topology vary. The simulations assume rather simplified models, and should be understood as an upper bound on the achievable gain in a practical system. The results obtained are valuable in filtering out settings with irrelevant trade-offs.
 G.P. Koudouridis, H.R. Karimi, K. Dimou, "Switched multi-radio transmission diversity in future access networks," in Proc. IEEE Semi-annual Vehicular Technology Conference, VTC-2005-Fall, vol.1, pp. 235- 239, 25-28 September 2005.
Summary: In this paper the exploitation of the multi-radio transmission diversity in a multi-cell environment is performed by means of packet scheduling algorithms. The packet scheduling is performed in two steps: (i) a user scheduling step, followed by a (ii) radio access allocation step that implements switched multi-radio transmission diversity. For the scheduling, different schemes are proposed and their performance is evaluated via simulations. Simulation results show significant average gains in spectral efficiency when MRTD is employed, as compared to scenarios where the radio accesses operate independently.
Contribution by author: The author of this thesis contributed in this paper as follows: (i) was the main author and wrote the paper with the support of the co-authors, (ii) contributed to the definition of the evaluation study and developed a simulator to evaluate the performance of MRTD, (iii) performed independent simulation experiments and contributed to the evaluation analysis of the simulations results, and (iii) contributed to the validation of the results based on a physical layer model implementation provided by the second author.
 H.R. Karimi, K. Dimou, G.P. Koudouridis, P. Karlsson, "Switched Multi-Radio Transmission Diversity for Non-Collocated Radio Accesses," in Proc. IEEE Semi-annual Vehicular Technology Conference, VTC2006-Spring, vol. 1, pp.167-171, 7-10 May 2006.
Summary: The study of this paper focuses on the switched MRTD employed across macro-cellular and pico-cellular radio accesses with non-collocated base stations. The evaluation is performed on the spectral efficiency gains when packets of data are jointly scheduled for downlink transmission over multiple independent radio accesses. Simulation results show that while significant gains can be achieved via MRTD among collocated macro-cell (or pico-cell) base stations, tight cooperation across non-collocated macro-cell and pico-cell base stations is only beneficial for a small subset of possible geometries. It has also been shown that the impact of channel quality indicator (CQI) reporting delays can be significant.
Contribution by author: The author of this thesis contributed in this paper as follows: (i) was a co-author contributing to all sections of the paper, (ii) contributed to the definition of the simulation study and performed independent simulation experiments for the evaluation analysis, and (iii) performed validation of simulation results provided by the first author.
 G.P. Koudouridis, P. Karlsson, "On the Performance of Multi-Radio ARQ and Packet Scheduling in Ambient Networks," in Proc. IEEE Semi-annual Vehicular Technology Conference, VTC2007Fall, pp.1456-1460, 30 September - 3 October 2007.
Summary: In this paper the focus is on the impact of the delays of measurements of radio link quality as reported by the radio interface. To alleviate degradations in spectral efficiency caused by reporting delays, an ARQ scheme utilising multiple RAs (MR-ARQ) in conjunction with MRTD has been devised and studied by means of simulations. Such multi-radio ARQ mechanism exploits the diversity across independent RAs, and results in gains in throughput over those achieved by RA-specific ARQ. MRTD when combined with MR-ARQ shows significant gains in spectral efficiency compared to the case where the radio accesses operate independently.
Contribution by author: The author of this thesis contributed in this paper as follows: (i) authored the paper with the review feedback of the co-author (ii) defined, developed and performed simulation experiments and analysed the simulations results.
 A. Yaver, G.P. Koudouridis, "Performance Evaluation of Multi-Radio Transmission Diversity: QoS Support for Delay Sensitive Services," in Proc. IEEE Semi-annual Vehicular Technology Conference. VTC2009-Spring, pp.1-5, 26-29 April 2009.
Summary: This paper extends the feasibility study to include service demands that are transmitted over established radio access technologies. Compared to the previous work this study differs in two ways: Firstly, the traffic model is based on an application with specific QoS demands and secondly the heterogeneous network scenario where the MRTD concept is applied is no longer based on physical layer abstraction but it utilises existing radio access technologies and protocols. More specifically, the network technologies in this study include HSDPA on UMTS and IEEE 802.11 and the performance of MRTD is based on delay sensitive services simulated by means of Voice-over-IP traffic. By applying various MRTD schemes, their impact and gains in performance have been observed in terms of average packet delay, packet loss and goodput as compared to a legacy system which operates under only one access for a given session.
Contribution by author: The author of this thesis contributed in this paper as follows: (i) supervised and assisted the first author to perform the study and co-authored the paper (ii) led the development of the simulator and the implementation of the radio access selection algorithm, (iii) defined the simulation experiments, which were developed and performed by the first author, and (iv) led the analysis of the simulation results.
Other related papers and contributions, not included in this thesis.
 G.P. Koudouridis, R. Agüero, E. Alexandri, J. Choque, K. Dimou, H.R. Karimi, H. Lederer, J. Sachs, R. Sigle, "Generic Link Layer Functionality for Multi-Radio Access Networks," in Proc. IST Mobile and Wireless Communications Summit, pp.1-5, 19-23 June 2005.
 H.R. Karimi, G.P. Koudouridis, K. Dimou, "On the Spectral Efficiency Gains of Switched Multi-Radio Transmission Diversity," in Proc. International Symposium on Wireless Personal Multimedia Communications, WPMC2005, pp.VII-1177-VII-1181, 18-22 September 2005.
 G.P. Koudouridis, R. Agüero, E. Alexandri, M. Berg, A. Bria, J. Gebert, L. Jorguseski, H.R. Karimi, I. Karla, P. Karlsson, J. Lundsjö, P. Magnusson, F. Meago, M. Prytz, J. Sachs, "Feasibility Studies and Architecture for Multi-Radio Access in Ambient Networks," in Proc. Wireless World Research Forum WWRF15, 8-9 December 2005.
 J. Sachs, R. Agüero, M. Berg, J. Gebert, L. Jorguseski, I. Karla, P. Karlsson, G.P. Koudouridis, J. Lundsjö, M. Prytz, O. Strandberg, "Migration of Existing Access Networks Towards Multi-Radio Access," in Proc. IEEE Semiannual Vehicular Technology Conference, VTC-2006 Fall, pp.1-5, 25-28 September 2006.
 G.P. Koudouridis, O. Queseth, "Research Challenges on Multi-radio Access Selection in Dynamically Composed Networks," in Proc. Swedish National Computer Networking Workshop, SNCNW2006, 26 October 2006.
 G.P. Koudouridis, R. Agüero, R.; Daoud, K.; Gebert, J.; Prytz, M.; Rinta-aho, T.; Sachs, J.; Tang, H.; , "Access Flow based Multi-Radio Access Connectivity," in Proc IEEE International Symposium on. Personal, Indoor and Mobile Radio Communications, PIMRC 2007, pp.1-5, 3-7 September 2007.
 J. Sachs, R. Agüero, K. Daoud, J. Gebert, G.P. Koudouridis, F. Meago, M. Prytz, T. Rinta-aho, H. Tang, "Generic Abstraction of Access Performance and Resources for Multi-Radio Access Management," in Proc. IST Mobile and Wireless Communications Summit, pp.1-5, 1-5 July 2007.
 A. Yaver, G.P. Koudouridis, "Utilization of Multi-Radio Access Networks for Video Streaming Services," in Proc. IEEE Wireless Communications and Networking Conference, WCNC2009, pp.1-6, 5-8 April 2009.
 G.P. Koudouridis, A. Yaver, M.U. Khattak, "Performance Evaluation of Multi-Radio Transmission Diversity for TCP Flows," in Proc. IEEE Semiannual Vehicular Technology Conference, VTC2009Spring, pp.1-5, 26-29April 2009.
Finally, selective parts of the material has been included in the Ambient Networks book  as chapter 8 about multi-radio access and paper  has been selected for publication in  as chapter 13:
 N. Niebert, A. Schieder, J. Zander and R. Hancock, Ambient Networks: Co-operative Mobile Networking for the Wireless World, John Wiley & Sons, Ltd, 2007, ISBN: 978-0-470-51092-6
 J. Sachs, R. Agüero, K. Daoud, J. Gebert, G. Koudouridis, F. Meago, M. Prytz, T. Rinta-aho and H. Tang, "Generic Abstraction of Access Performance and Resources for Multi-Radio Access Management", in Advances in Mobile and Wireless Communications, Views of the 16th IST Mobile and Wireless Communication Summit, F. István, J. Bitó and P. Bakki (Eds.), Lecture Notes in Electrical Engineering, 2008, Volume 16, Part IV, 239-259, doi: 10.1007/978-3-540-79041-9_13.
The thesis consists of two parts. The first part contains a description of the MRA architecture and the basic mechanisms that enable simultaneous utilisation of multiple radio accesses. These descriptions are given in Chapter 2 and Chapter 3. In particular Chapter 2 deals with the concept of GLL and the inter-layer and cross-layer integration of RATs. The GLL functions at different levels of integration are briefly described as well as the interactions between MRRM and GLL. In Chapter 3 the MRTD concept is defined and the various ways it can be applied are presented. The basic functions required for implementing MRTD are also described in more detail. These descriptions provide the conceptual framework required to understand the concluding summaries of the results from the evaluation studies in Chapter 4. In Chapter 5, the thesis is summarised and recommendations for future work are outlined. In the second part, the papers that constitute the contributions of this thesis are reprinted in verbatim.