A Network-Assisted Approach for RAT Selection in Heterogeneous Cellular Networks

Abstract

When several radio access technologies (e.g., HSPA, LTE, WiFi and WiMAX) cover the same region, deciding to which one mobiles connect is known as the Radio Access Technology (RAT) selection problem. To reduce network signaling and processing load, decisions are generally delegated to mobile users. Mobile users aim to selfishly maximize their utility. However, as they do not cooperate, their decisions may lead to performance inefficiency. In this paper, to overcome this limitation, we propose a network-assisted approach. The network provides information for the mobiles to make more accurate decisions. By appropriately tuning network information, user decisions are globally expected to meet operator objectives, avoiding undesirable network states. Deriving network information is formulated as a Semi-Markov Decision Process (SMDP), and optimal policies are computed using the Policy Iteration algorithm. Also, and since network parameters may not be easily obtained, a reinforcement learning approach is introduced to derive what to signal to mobiles. The performances of optimal, learning-based, and heuristic policies, such as blocking probability and average throughput, are analyzed. When tuning thresholds are pertinently set, our heuristic achieves performance very close to the optimal solution. Moreover, although it provides lower performance, our learning-based algorithm has the crucial advantage of requiring no prior parameterization. Index Terms-Radio access technology selection, semi-Markov decision process, reinforcement learning, heterogeneous cellular networks. I. INTRODUCTION T HE demand for high-quality and high-capacity radio networks is continuously increasing. It has been reported that global mobile data traffic grew by 81 percent in 2013 [1]. Furthermore, monthly mobile traffic is forecast to surpass 15 exabytes by 2018, nearly 10 times more than in 2013 [1]. Along with this impressive growth, mobile operators are urged to intelligently invest in network infrastructure. They also need to reconsider their flat-rate pricing models [2], seeking positive return-on-investment.

Intelligent Reward-Based Data Offloading in Next-Generation Vehicular Networks

“…When there is high load in WiFi network, the throughput may degrade and the data users need to connect to LTE networks. Policy Iteration and SMDP Q-Learning algorithms are proposed in [14] to solve the SMDP problem. In [36], the Discrete-Time Markovian Decision Process (DT-MDP) is introduced along with the Centralized Q-Learning (QC-learning) to optimize the energy consumption by Heterogeneous Cellular Network (HCN).…”

“…Fig. 9 shows the RSU Utilization in percentage (%) for various offloading approaches [14] including the proposed IR-DON method where there is high and low load in the network with respect to priority of QoS. In the proposed IR-DON architecture, maximum RSUs are utilized by the user and almost all users get dedicated QoS connection.…”

Improved network selection for multimedia applications

“…So this method does not give optimal result for large number of users. A network‐assisted approach for selection of network in heterogeneous cellular networks jointly enhance network performance and user experience. In this approach, a semi‐Markov decision process is used, and it is solved by policy iteration algorithm.…”

Vehicular HetNet load distribution using centralized control of software‐defined network

“…User's individual performance‐based VHO approaches are available, but network performance improvement centric methods are not available in the literatures . A multiple base station‐based approach using multimode terminals (MMT) are seen in literatures which is capable of parallel transmission too.…”