Energy Efficient
6.1.3 Handover Decision Algorithms
6.1.3.3 Cost-function Based
This class includes algorithms that integrate a wide range of HO decision parameters within a single cost-function. The outcome of the cost-function is subsequently used as the primary HO decision criterion. Even though the use of cost-functions enables more sophisticated HO decision making, the weights related to the cost-function should be carefully adapted as well. The signaling overhead for assessing the parameters should also be investigated. The remainder of this section discusses three representative algorithms of this class.
a. Cost-function based HO Algorithm for the Macrocell-Femtocell Network
A cost-function based algorithm is proposed in [62] aiming to enhance inbound mobility to femtocells. The algorithm applies to the single-macrocell single- femtocell scenario, where the UE is connected to a macrocell and enters the coverage of a femtocell. The cost-function can be based on either the RSRP or the RSRQ status of the macrocell and femtocell stations. To ease the reader, we indicate the serving macrocell with and the target femtocell with . For a cell
∈ { , }, the cost-function is denoted by ( ) and is given as follows: ( ) = ( )∙ ( )∙ ( )
∙ ( ) ( ) (6.3) where ( ) denotes the RSRP or RSRQ status of cell , ( ) the capacity of cell , ( ) a factor used to adjust the value of the cost-function, ( ) a factor adapted with respect to the type of cell , and ( ) the number of UEs camped on cell .
Figure 62: Zhang et al. cost-function based HO algorithm [62]
The proposed algorithm is illustrated in Figure 62. Firstly, the algorithm validates the UE membership status on the target femtocell. The HO decision parameter set is subsequently acquired and the cost-functions for the serving macrocell and the target femtocell are evaluated. The HO decision phase continues either if the cost-function outcome for the serving macrocell is lower than a prescribed threshold, denoted by , or if the cost-function outcome for the target femtocell is greater than the one of the serving macrocell plus a HHM. If the UE moves faster than 30 km/h the algorithm avoids inbound mobility to the
target femtocell. In the opposite case, the algorithm continues either if a) the UE speed is lower than 15 km/h or b) the UE speed is between 15 km/h and 30 km/h and the user connection is of real-time. In both cases, the proposed algorithm waits for a time interval equal to and evaluates whether the cost-function outcome for the femtocell is greater than the one of the macrocell plus a HHM. If so, the bandwidth availability on the femtocell is validated and a HO is initiated.
The proposed algorithm accounts for the bandwidth availability and the UE speed to lower the HO failure probability due to admission control and mitigate the number of unnecessary HOs for medium to high speed users, respectively. The integration of the number of camped UEs within the cost-function enables load balancing between the macrocell and the femtocell tiers as well. However, even though the ( ) and ( ) parameters majorly affect the decision outcome, a methodology for optimizing their values is not provided. Moreover, the motivation for using the specific speed thresholds is not discussed, whereas the signaling and delay overhead for acquiring the number of camped UEs and the available bandwidth of the target cell need to be further examined.
b. User State and Signal Quality based HO Algorithm for Hierarchical Cell Networks
The authors in [63] describe a cost-function based algorithm that applies to the single-macrocell single-femtocell scenario, where the UE receives service from a macrocell and enters the coverage of a femtocell. The algorithm uses a simple cost-function, which integrates the traffic-type and the UE speed. Two traffic types are considered: voice/video and data. To ease the reader, we indicate the serving macrocell with and the target femtocell with .
The proposed algorithm is depicted in Figure 63. Upon HO decision triggering, the proposed algorithm acquires the RSQ of the target cells, assesses the UE speed and examines the traffic-type of the user connection. The algorithm subsequently compares the RSQ of the macrocell and femtocell stations. Note that we replace the SINR comparison of the original manuscript with the RSRQ comparison, given that the SINR of the target femtocell can be only evaluated upon service reception. The algorithm subsequently compares the UE speed with an absolute speed threshold and classifies the traffic-type of the user to calculate the speed and traffic-type costs, denoted by ( ) and ( ), respectively. These costs are integrated in the cost-function, denoted by , which is ultimately used to reach to the HO decision.
The proposed algorithm integrates the UE speed and traffic-type within a simple cost-function. The use of these parameters is expected to lower the HO probability for medium to high speed users and enhance the QoE of the mobile users. However, even though the incorporation of the RSQ parameters aims at improving the SINR performance at the UEs, the absence of a HHM during the relative RSQ comparison may unpredictably raise the HO probability due to fast variations of the wireless medium. The performance of the proposed algorithm should also be compared to other non-RSQ based algorithms and evaluated in terms of interference, throughput and energy consumption at the UEs.
Figure 63: Xu et al cost-function based HO algorithm [63]
c. Cost-Based Adaptive HO Hysteresis Algorithm to Minimize the Handover Failure Rate
The proposal in [64] is a representative cost-function based algorithm that readily applies to the two-tier macrocell-femtocell network. The cost-function consists of a weighted summation of parameters related to the UE speed, cell load, and number of user connections. The cost-function outcome is integrated in a standard RSS-based procedure as an adaptive HHM. The algorithm uses individual cost-functions for the cell load, the UE speed, and the number of UE connections, denoted by , , and , respectively. The corresponding weights are denoted by , , and , respectively, whereas the cost-function by , , .
The proposed algorithm is depicted in Figure 64. Upon HO decision triggering, the algorithm calculates a) the number of real-time and non real-time connections of the tagged UE, denoted by and , respectively, b) the UE speed, and c) the cell load of the current serving and the candidate cells, denoted by ( ) and ( ), respectively. The algorithm subsequently evaluates the cost-functions for the cell load, the UE speed and the number of user connections. The outcome
is integrated within the cost-function , , . Note that the UE speed is normalized to the maximum speed among the UEs, while the ( ) and ( ) parameters are expressed as the ratio of the occupied bandwidth to the total bandwidth of the cell. The result of the cost-function is multiplied with an adjustment parameter, denoted by , and is incorporated in a standard RSS-based procedure as an additional HHM.
Figure 64: Lee et al. HO algorithm [64]
Among the strong features of the proposed algorithm is the use of the adaptive HHM, which can be readily integrated in the standard RSS-based HO decision. The incorporation of the load difference between the serving and the target cell is also expected to balance the load distribution among the cells and increase the bandwidth availability for the UEs. A reduced HO probability is expected for medium to high speed users as well, given that the algorithm accounts for the UE speed. However, the proposed algorithm lacks of a methodology for optimizing the cost-function weights. On the other hand, even though the proposed algorithm is shown to attain a lower HO failure rate compared to other competing algorithms, further numerical results are required to evaluate its performance in terms of HO probability, load distribution and throughput. Increased signaling is also required to commute the load occupation of the target cells to the serving cell.