Performance Evaluations

For simplicity, the CRFES planning defines that the problem of searching for only energy-efficient Wi-Fi spots on the AP maps. The problem is that finding a path subject to multiple constraints is inherently hard. This is why we assume the foregoing condition. This assumption is acceptable as the search area is limited. Note that our approach is heuristic and there is no guarantee that CRFES shows the best route.

3.5.2.2 CRFES and SR Characteristics

The energy consumption for CRFES and SR are reflected in the total ER. This value is calculated using the ER for the transmission and idle states. In the case of CRFES, the transmission ER is low because the user only transmits some data at the energy-efficient spot. If this spot is far from the SR, the idle ER is high. In the case of SR, the transmission ER is high. The user primarily communicates using cellular networks because Wi-Fi coverage is narrow.

Chapter 3: A High-QoS Communication Route Navigation for Mobile Users

72

3.5.2.3 Evaluation Metrics

To evaluate the performance of CRFES, we employ two evaluation metrics:

Longcut Ratio (LR) for the user cost, and Energy Saving Ratio (ESR) for the gain of CRFES. These metrics are calculated as follows:

[sec]　

SR a for time Total

[sec]

CRFES a

for time Total



LR (3.14)

SR a for ER Total

CRFES a

for ER Total

ESR  (3.15)

3.5.2.4 Evaluation Models

We evaluate CRFES performance using computer simulations and in a real environment. The topology of our simulation maps is an 80 × 80 square lattice, with two adjacent vertices being 5 m apart. The mobile user moves from a start spot (0, 60) to a goal spot (74, 60) at 1 m/s. The shortest time is 370 s. The user can communicate with the Wi-Fi/cellular networks while moving. Our maps had 10 APs and 7 cellular spots.

We assume that the Wi-Fi APs have the same energy consumption characteristics as shown in Figure 3.32. At the cellular spots, the available throughput within 50 m of the cellular spots is uniform, and the energy consumption characteristics are same as shown in Figure 3.33. The energy consumption beyond 50 m is also uniform on average. The locations of the Wi-Fi/cellular spots are randomly placed on each map. We provide 1000 different maps. As shown in Figure 3.34, one of them calls the Shinjuku map refers to Shinjuku, one of the major cities in Japan.

We also evaluate the CRFES performance in a real environment, Shinjuku. We search the CRFES and SR using Google MAPs APIs. The evaluation map is shown in Figure 3.35. In this evaluation, although we can use several served Wi-Fi spots, we only use the FREESPOT APs [63]. We select Carrier A (LTE) as the cellular network. A TCP flow is transmitted using iPerf with the Galaxy S2 LTE SC-03 device. The available throughput at the most energy-efficient spot is 11 Mbps, which is roughly equal to that obtained in the simulation. In both evaluations, the amount of transmission data of

Chapter 3: A High-QoS Communication Route Navigation for Mobile Users

73

CRFES is the same as that of SR. According to the general user behaviors, the SR user constantly communicates with the Wi-Fi/cellular networks.

Figure 3.34: One of the examples using CRFES and an energy consumption in Shinjuku map. Blue area indicates energy efficient area [8].

Figure 3.35: Evaluation map for real environment in Shinjuku [8].

Chapter 3: A High-QoS Communication Route Navigation for Mobile Users

74

3.5.2.5 LR and ESR Characteristics

The LR and ESR depend on the location of the most energy-efficient Wi-Fi spot. The LR gets increase as the energy-efficient spot gets farther away from the SR.

Although the total ER of CRFES gets higher, it is lower than that of SR. This is because the CRFES user’s device mainly consumes energy in the idle states. When the LR gets one, which indicates that the energy-efficient spot is located on the SR, the CRFES shows the best performance.

Because results for the other scenarios have similar characteristics, the results of LR and ESR for Carrier A (LTE) are only shown in Figure 3.36. The characteristics of throughput and energy consumption between 3G and LTE are approximately the same in our experiments. Figure 3.36 is sorted according to the LRs. From this figure, the CRFES succeeds in significantly extending the device’s battery life in all maps, and the LRs are within 1.5 in approximately 70% of maps. This indicates that the CRFES user can obtain longer battery life at low cost.

Figure 3.36: Results for LR and ESR in 1000 different maps (The Shinjuku map is located in map ID 649.)

The results for the Shinjuku map are shown in Figure 3.37. The simulation results form a part of Figure 3.36. Figure 3.37 also shows the gain of CRFES in the real environment. As shown in Figure 3.37, both ESR and LR results in the real environment

Chapter 3: A High-QoS Communication Route Navigation for Mobile Users

75

are higher than those results in the simulation. This is primarily because of the fact that the most energy-efficient spot is located on the fifth floor of a building. Therefore, the idle time is longer.

Figure 3.37: Comparison of time cost and energy saving ratio between the real environment and simulation in Shinjuku [8].