3.5 Predictive Motion and Interference-based Scheduling (PMIS) Algorithm
3.5.2 PMIS Algorithm
Combining the PIS algorithm and the aforementioned motion predication mechanism, the PMIS algorithm is proposed. Similarly, let L = {𝑁𝑇, 𝑁𝑅} denotes a direct communication pair, where 𝑁𝑇 and 𝑁𝑅are the transmitter and receiver respectively. Fig. 3.5 illustrates the flowchart of the PMIS algorithm, which consists of nine important steps. It can be observed that Steps 1 to 4 is the four major steps of the PIS algorithm (as mentioned in Section 3.4), which is utilized to arrange a proper communication period for a pair of MSs that are expected to conduct direct communication. The additional steps, including Steps 5 to 9, of the PMIS algorithm is detailed as follows.
Step 5: In this step, the future positions of 𝑁𝑇 and 𝑁𝑅are estimated based on (3.18).
Step 6: Based on the estimated position of 𝑁𝑇, the possibility of direct communication for L in the next DL subframe is calculated in this step. The DL feasible region for L is utilized to determine the calculating result. If the result shows that L can be arranged in the next DL subframe, the algorithm will be complete. Otherwise, it will go to Step 7.
Step 7: This step is utilized to predict whether the direct communication between 𝑁𝑇 and
L
Begion of
Figure 3.5: Flow diagram of PMIS Algorithm.
𝐷𝑁𝑁𝐵
𝑅, the DL feasible region is acquired to predict the availability of direct communication for L. If the communication is permitted, the algorithm will execute Step 9. Otherwise, it will go to Step 8.
Step 8: Let 𝛿(𝑁𝑅) denotes the number of times to predict 𝑁𝑅’s position and 𝜉 is a pre-defined threshold value. With the increment of 𝛿(𝑁𝑅), the predicted error for 𝑁𝑅’s position may also be raised. In this step, therefore, 𝛿(𝑁𝑅) is examined. If 𝛿(𝑁𝑅) > 𝜉, the BS will request 𝑁𝑅to provide its actual position information. Otherwise, it will go to Step 9.
Step 9: The functionality of this step is similar to that of Step 8, but the targeted MS is 𝑁𝑇 instead of 𝑁𝑅. Let 𝛿(𝑁𝑇) denotes the number of times to predict 𝑁𝑇’s position. If 𝛿(𝑁𝑅) > 𝜉, the BS will request 𝑁𝑇 to provide its actual position information and complete the algorithm. Otherwise, the algorithm will directly be ended.
3.6 Performance Evaluation
In this section, simulations are conducted to evaluate the performance of the proposed PIS and PMIS algorithms on the adaptive point-to-point communication (APC) approach (as mentioned in Chapter 2). A 19 cell-based wrap around topology [18] is considered as the simulation layout. Each cell consists of a centered BS and various number of uniformly dis-tributed MSs. Both the intra-cell and inter-cell traffic flows are considered in the simulation, wherein each MS generates 20 traffic flows with randomly selected destination. The packet ar-rival process of each flow is assumed to follow a Poisson process with rate 𝜆 = 2 packet/frame.
The size of each packet is selected to follow the exponential distribution with the mean value of 200 bytes. Since the scheduling algorithm is not specified in the IEEE 802.16 standard, the deficit round robin (DRR) [19] and weighted round robin (WRR) [20] algorithms are selected as the BS’s DL and UL schedulers respectively. The DRR algorithm is also utilized by the SS to share the UL grants that are provided by the BS for its corresponding UL connections.
The simulation is implemented via MATLAB even-driven simulator. Each obtained result is average from 100 simulation runs. The parameters adopted within the simulation are listed in Table 3.1.
Table 3.1: Simulation parameters for PMIS algorithm
Parameter Value
OFDM symbol duration 34 𝜇s
Maps modulation BPSK
Data modulation QPSK, 16-QAM, 64-QAM
Frame duration 10 ms
Number of MSs 10, 20, 30, 40, 50 Mobility model random waypoint mobility Mean velocity ( ¯𝑉 ) 0, 5, 10, 15, 20 m/s
Figure 3.6: Performance comparison of user throughput and control overhead versus number of MSs (with ¯𝑉 = 10).
Fig. 3.6 illustrates the performance of the user throughput and control overhead under various number of MSs for the compared schemes. The mean velocity of 10 m/s is considered for all MSs in this comparison. As can be expected that the user throughput increases as the number of MSs is augmented for all the schemes. Due to the effect of direct communication conducted among MSs, the performance of the APC-based approaches outperform that of the conventional indirect scheme (denoted as IEEE 802.16). For the original APC approach (i.e.,without the consideration of scheduling algorithm for directly communicable pairs), it is observed that conducting direct transmissions in UL subframes (denoted as APC-UL) has better performance than that in DL subframes (denoted as APC-DL). The reason can be contributed to the potential inter-cell interferences to neighbor-cell MSs that are introduced by conducting direct communication in DL subframes, which can consequently decrease the entire system performance. If the direct transmissions are arranged in UL subframes, the con-ventionally indirect communication pairs will not be interfered since all the MSs are served as transmitters during those UL subframes. With the consideration of scheduling algorithms for directly communicable pairs in the APC approach (i.e., the APC-PIS and APC-PMIS schemes), the enhanced performance of user throughput is acquired. It is because that the directly communicable pairs are properly arranged in either DL subframes or UL subframes without inducing additional interference in the proposed PIS and PMIS algorithms. Compar-ing the APC-PIS and APC-PMIS approaches, it is observed that the higher user throughput and control overhead are shown in the APC-PIS scheme, which can be attributed to the acquirement of MSs position information. In the APC-PIS scheme, the exact position infor-mation of MSs are acquired via the updates of MSs, which results in large amount of control overhead. On the other hand, the position information of MSs are properly estimated by the BS in the APC-PMIS scheme, which efficiently reduces the necessary of position updates by MSs. Consequently, the lower control overhead is shown in the APC-PMIS approach in comparison with the APC-PIS scheme. Noted that the control overhead means the overhead related to the position updates of MSs. Since the position information of MSs is not utilized in the APC-DL, APC-UL, and IEEE 802.16 approaches, there is no control overhead in these
0 5 10 15 20
Figure 3.7: Performance comparison of user throughput and control overhead versus mean velocity ( ¯𝑉 ).
schemes.
Performance comparison with an increasing velocity ranging from 0 to 20 m/s is shown in Fig. 3.7, wherein 20 MSs are considered. As can be expected that the user through-put decreases as the mean velocity of MSs is increased for all the schemes excepted for the IEEE 802.16 approach, which can be attributed to the effect of direct communication. With larger velocity for the MSs, the variation of distances and channel conditions among MSs are changed frequently, which reduces the average life time of direct links. In such a case, the communication operation for the original direct communication pairs are changed from the direct manner to the indirect manner, which results in the decrement of user throughput. Due to the same reasons addressed for Fig. 3.6, the APC-PIS approach has the best performance of user throughput and worst control overhead among the compared schemes.
3.7 Concluding Remarks
In this work, a scheduling algorithm for each pair of MSs that is expected to conduct direct communication is proposed. Both the interference region and feasible region for the pair of
MSs to perform direct communication are studied and calculated. Based on these two types of information, the predictive interference-based scheduling (PIS) algorithm properly arranges the MSs to conduct direct communication in either DL subframe or UL subframe without increasing additional interference for other communications. Furthermore, since MSs may move around, a motion prediction mechanism is proposed. Combining the predication mech-anism with the PIS algorithm, a predictive motion and interference-based scheduling (PMIS) algorithm is given to reduce the control overhead regarding the updates of MSs positions.
The efficiency of the proposed PIS and PMIS algorithms are evaluated and compared via simulations. Simulation studies show that both of the PIS and PMIS algorithms efficiently enhance the performance of user throughput in comparison with the original adaptive point-to-point (APC) approach. Furthermore, the control overhead resulted from the PIS algorithm is significantly reduced by the PMIS approach.