Rapid advances in wireless access technologies and in-network processing can signif- icantly assist in implementing smart healthcare systems through providing seamless integration of heterogeneous wireless networks, medical devices, and ubiquitous access to data. S-health demand for high data rates and QoS has motivated us to leverage the development of cellular networks into dense heterogeneous networks (HetNets) with the utilization of multi-Radio Access Technology (RAT). It is es- sential for each user/device to leverage different RATs, hence, the available radio resources across different spectral bands, to communicate with the network infrastruc- ture [135]. Utilization of the spectrum across diverse radio technologies is expected to significantly enhance network capacity and QoS for emerging applications such as remote healthcare monitoring. However, this imposes an essential need to develop innovative networks association mechanisms that account for energy efficiency while meeting application quality requirements.
In accordance with the new trends foreseen for 5G systems, this chapter proposes an efficient networks association mechanism with adaptive data compression for en-
hancing the performance of s-health systems.1 Different performance matrices have been considered, in addition to networks characteristics and application requirements, in order to find an efficient solution that grasps the conflicting nature of the various users’ objectives and addresses their inherent tradeoffs. The proposed mechanism adopts a user-centric approach towards exploiting heterogeneous wireless networks to optimize medical data delivery over heterogeneous s-health systems. In particular, we focus on answering the following questions:
1. Which network(s) should be selected among multiple Radio Access Networks (RANs)?
2. What is the optimal level of data compression to be used?
3. What is the amount of data that should be sent through each selected RAN after compression?
While addressing the above issues, we account for both network characteristics and application requirements, providing a solution, which achieves an optimal energy- quality-cost tradeoff.
5.2
Related Work
With the rising tendency toward network densification, various radio technologies, such as 3G, 4G and WiFi, could be jointly leveraged to enable seamless connection to users with high levels of quality of experience [137]. Simultaneously utilizing of multiple radio technologies turns to be even more serious as the user demand and QoS requirements proliferate, while the available wireless resources remain limited. Consequently, the upcoming 5G systems are expected to have dense and irregular HetNets, where the user will be able to access the system through different points of access. Thus, it is crucial to develop techniques that can efficiently leverage the available radio resources across different spectral bands using multi-RAT [108].
The association with network infrastructure may be concurrent, exploiting the multihoming feature of mobile devices to establish simultaneous associations with different access networks, or switch from one point of access point to another, within
5.2 Related Work 103
the same RAN or across different RANs. In both cases, several schemes have been proposed in the literature for network selection and association in HetNets. The proposed approaches can be broadly classified into four categories: cost-function based, decision making processes using game theory, Markov decision processes (MDPs), and optimization based. Cost function-based schemes proactively select the network with the highest/lowest utility or cost function [127][138]. Although the approach can achieve a near optimal solution, it is often hard to prove it. According to the game theory approach, instead, users in different service areas compete for the bandwidth offered by different wireless networks [139][140]. The resulting algorithms are, in general, complexity-prohibitive, and their convergence is not guaranteed. Even in case of convergence, they do not necessarily converge to an optimal solution. MDPs have been used also to study network switching between different RATs [141][142]. However, finding the optimal solutions is again cumber- some, especially in the case of large networks [137]. Formulating network selection problem as an optimization problem with low or moderate complexity is also not a trivial task. Finding optimal resource allocation and user association, subject to resources and/or power constraints, may result in an NP-hard problem [143]. One way to make the problem tractable is by using constraints relaxation or variables transformation, or by envisioning online adaptive methods such as Q-learning [144].
Finally, the existing work on concurrent association mainly focus on designing a traffic scheduler over different device interfaces considering users incentives for collaboration and bandwidth sharing [145], a transport-layer control protocol to en- able concurrent multipath transport [146], or content-aware transport-layer protocols [147]. Other studies have focused on the resource allocation problem for parallel transmission considering multi-RATs [148][149]. For instance, a sub-optimal solu- tion is presented in [148] by utilizing the intrinsic quasi-concavity of the formulated problem. While, in [149], the authors develop a framework of multi-RAT system, where a small cell serves a number of mobile users via IEEE 802.11 WLAN and 3GPP LTE access technologies. A scheduler at the small cell is developed in order to minimize the total transmission power subject to QoS constraints on the users’ transmission rates. In [150], an urban deployment scenario is investigated, where WiFi small cells are overlaid on top of the 3GPP LTE network. The authors present user-centric network selection algorithms to minimize feedback overhead while considering user’s preferences. A good review of the mathematical methods that are applied to the network selection problem, including cost-function, game theory,
multiple attribute decision making, combinatorial optimization, fuzzy logic, and Markov chain, can be found in [151].
To the best of our knowledge, none of the aforementioned work advocates a user-centric approach for efficient networks association taking into consideration context-aware in-network processing to optimize the delivery, cost, and latency of the medical data. Thus, our main contributions are summed up as follows:
1. A multi-objective optimization problem is formulated that enables each PDA to optimally set its data compression ratio and select the RAN(s) for data transmission, in an energy-efficient and cost-aware manner while ensuring an acceptable signal distortion.
2. We propose an analytical solution for the formulated optimization problem, by decomposing it into two sub-optimizations. The two sub-problems turn out to be solvable with low complexity, and they are analytically proved to lead to the same optimal solution as the original problem.
3. We design a distributed, iterative, PDA-centric algorithm considering the case where initially PDAs have just a rough estimate of their resource share on the available RANs. Then, the convergence behavior of the proposed algorithm to the optimal solution is analyzed and proved analytically.
4. Finally, the performance of the proposed approach is evaluated and compared against that of state-of-the-art techniques. Our results demonstrate that the proposed solution allows for high-quality healthcare monitoring of patients, it significantly outperforms other solutions, and swiftly adapts to varying network dynamics.