The incentive scheme is compared with three conventional delivery schemes as follows.
Epidemic  In this scheme, bundles are flooded when a bundle carrier encounters other nodes that do not possess a copy of the bundle.
Direct Deliver  In this scheme, the source holds the bundle until it comes in contact with the destination.
PRoPHET +  In this scheme, the carrier forwards the bundle to other nodes with the weighted function determined by the nodes buffer size, power, and predictability.
Figure4.3 shows the delivery ratio by comparing the proposed scheme with other three schemes. In Fig. 4.3, it is shown that the proposed scheme outperforms other existing delivery schemes when the number of selfish nodes in MSNs changes. In the epidemic scheme, since there is no incentive strategy, selfish nodes refuse to relay bundles. Therefore, it causes that many bundles are dropped when bundles are expired or the buffer of the node is overflowing. In PRoPHET+, the bundles cannot be delivered to the destination with a high delivery ratio due to the selfish nodes. For the direct-delivery scheme, it shows the worst performance since it only delivers the bundle when arriving at the destination.
Fig. 4.3 Delivery ratio versus percentage of selfish nodes
Fig. 4.4 Delivery delay versus percentage of selfish nodes
Fig. 4.5 Delivery ratio versus buffer size
Figure4.4 shows the comparison of delivery delay. From this figure, we can infer that the proposed incentive scheme has the lowest delay as almost all nodes participate in bundle delivery. In epidemic, because some nodes are still selfish, bundle owners are required to wait for cooperative nodes. In PRoPHET+, the carriers have to wait for the cooperation nodes for bundle forwarding, thus resulting in longer delay. For direct-delivery, it takes a long time to encounter the destination, thus resulting in the largest delay.
Next, we test the delivery ratio and delay with different buffer sizes, where the buffer size is changed from 20 to 50 MB, and the percentage of selfish nodes is fixed to 0.6. In Fig. 4.5, although all of the delivery ratios of the three schemes increase when the buffer size of each node is increased, the proposed scheme has the maximum delivery ratio. In the epidemic scheme, bundles are stored in the buffer of a node for a long time, thus resulting in the fact that many bundles cannot be forwarded when the buffer is overflowing. In PRoPHET+, as many bundles may be dropped when
Fig. 4.6 Delivery delay versus buffer size
the limited buffer of each node is full, the delivery ratio cannot be improved much. In the direct-delivery scheme, due to the lack of cooperation of nodes, bundles can be forwarded only when the source node encounters the destination node. Therefore, the delivery ratio is the lowest and is almost unchanged with the increase in buffer size. For the proposed incentive scheme, mobile nodes are willing to forward bundles through bargaining, although each node has a small-sized buffer.
Figure4.6 shows the delivery delay when the buffer size of each node is changed. From Fig.4.6, we can observe that the delivery delay of each scheme decreases with the increase in buffer size. In the epidemic scheme, mobile nodes have to wait for the cooperative nodes to forward bundles. Moreover, due to the small buffer size, some nodes cannot receive bundles or drop old bundles when the buffer is full. As a result, it takes a long time to forward bundles to the destination node. In PRoPHET+, due to the limited buffer, the node drops old bundles when the buffer is full, where the delivery delay cannot be efficiently reduced. In the direct-delivery scheme, the forwarding of a bundle depends on the encounter between the source node and the destination node, which takes a long time. Compared with the other three schemes, the delivery delay of the proposed scheme is the minimum. In the proposed scheme, although the buffer size is limited, the mobile nodes are willing to forward bundles with an appropriate transaction price, resulting in the fact that the bundle can reach the destination quickly.
The experiments have shown that the proposed incentive scheme can achieve the largest delivery ratio and the lowest delivery delay, compared with other existing protocols when the number of selfish nodes is changed. In addition, when the buffer of each node is changed, our incentive scheme can also obtain better performances in term of both the delivery ratio and delivery delay than others. According to these results, our proposed scheme can outperform other existing algorithms.