LQER: A Link Quality Estimation based Routing for Wireless Sensor Networks

1. Introduction

2. Related Work

3. Link Quality Estimation Routing Protocol Design

3.2. Minimum Hop Field Establishment

MHFR can provide optimal path to send data to the sink. In WSNs, let the hop count of the sink node be 0. For other nodes, the minimum hop count is defined as the number of intermediate nodes from that node to the sink on the optimal path plus 1. We use the flooding algorithm to establish the hop count field as shown in Table 1. Initially, each node sets its hop count to, for example 1000, more than the maximum hop count reachable in the network. After the sink broadcast an ADV (advertisement) message containing its own hop count (0 initially), the message propagates through the network. Upon hearing the an ADV message from node M, node N has a path with hop count H_M + 1, where H_M is the hop count of node M. Node N then compares its current hop count H_N with H_M + 1. If the new hop count is smaller, it sets H_N to H_M + 1 and broadcasts an ADV message with its new hop count. If the new hop count equals to its current one, it adds node M to its forwarding node set but does not broadcast any ADV message. If the new hop count is larger, it just rejects the message data. Eventually, every node may calculate the minimum hop to the sink through flooding and get each own forwarding node set.

Table 1. Power Model of MICA2 Node With Sensor Board in WSNSim

Suppose the message forwarding time of each node is the same T, it can be proved that however long the node is apart from the sink, it will hear a message with minimum hop only once [1]. However, in above design, some practical issues have been ignored such as delays from transmission, propagation, processing and channel error. For example, for a message with 30 bytes, if the transmission rate is 9600bps, the transmission time is $\frac{30\times 8}{9600}=0.025\text{s}$. If the delays are near to or lager than transmission time, some nodes may receive more than one message containing minimum hop count. To reduce the total number of messages in the network, we introduce a random delay T_w. When the node hears a message containing smaller hop count, it defers the broadcast to the time T_w. In that case, it may broadcast only once, carrying its minimum hop count.

Algorithm 1: Minimum Hop Field Establishment Algorithm

node N receives a message

if the received hop count < current hop count then

update current hop count;

add the sender to the forwarding node;

add the sender to the forwarding node

else if received hop count = current hop count then

add the sender to the forwarding node set

else

reject the message data

end

end

3.3. Link Quality Based Routing Protocol

LQER protocol integrates the approach of minimum hop count field and (m, k), which makes the routing energy-aware and low loss rate. Thus the whole network may obtain a longer lifetime and a better link quality.

Algorithm 2: Link Quality Table Maintenance Algorithm

choose the path and transmit the data packet

update the information in the link table

if data packet successful transmitted then

shift all bits in k-sequence towards left and add 1 to the rightmost

else

do shift and add 0 to the rightmost

end

end

Algorithm 3: Link Quality Estimation Routing Algorithm

1: receive the routing data

2: list all the nodes that are 1 hop count less than the current node

3: choose the node in the list that have largest value of$\frac{m}{k}$

4: broadcast routing data

5: finish

The algorithm of link quality table maintenance and link quality estimation based routing is described in Table 2 and Table 3. We take the example in Figure 1 to illustrate the LQER. Suppose minimum hop field has been established and the minimum hop count of node S is 10. Its forwarding node set includes A, B, C and D with the same hop count 9. The forwarding node set of node B includes E, F and G with hop counts 8. When node S has data packet to send to the sink, it first chooses the node from its forwarding set, the one with the largest value of $\frac{m}{k}$, which means the best link quality in history. Not like MHFR, the forwarding node is selected randomly from forwarding set. In this example, node B has the largest value of $\frac{m}{k}$ and is chosen to forward the packet. The same way is adopted by node B to choose its forwarding nodes. Obviously, node F is the victor. In this way, all the data packets can be forwarded through the shortest and most effectively reliable path and will not generate any extra information. When a node is forwarding the data, all the neighbors with 1 hop less can hear the message. But only the node designated in the message will forward the data. Thus the number of nodes participating in data forwarding is the least and the energy consumption is optimized.

Figure 1. Routing Tree Example

4. Simulation and Evaluation

4.1. Simulation Environment: WSNSim

WSNSim developed by Lin et al is a component based, event driven runtime and mote power modeling simulation environment to simulate the energy usage in each node for certain applications [22][23]. The components in WSNSim are similar to Mote developed by University of California at Berkeley, which include CLOCK, SENSOR, ADC, LED, RADIO and APPLICATION. The function of each component is defined as follows.

-CLOCK:

in charge of timing, can offer the current simulating time, is the basic of the simulating events.

-SENSOR:

provides the sensor data according to the requirement in the application.

-ADC:

collects the SENSOR data.

-LED:

shows the status of the node.

-RADIO:

communicates with other nodes and base station.

-APPLICATION:

performs specific application.

The power model used in WSNSim is from Mote MICA2 node with sensor board in a 3V power supply. Table 1 presents the power model for the Mica2 hardware platform. As the table shows, the different CPU power modes cover a wide range of current level, from 103μA in the power-down state up to 8mA when actively executing instructions. Likewise, the choice of radio transmission power affects current consumption considerably, from 3.7mA at -20dB to 21.5mA at +10dBm. However, in many of our applications the radio is almost always listening for incoming messages, which consumes 8mA regardless of transmission activity.

In our simulation, it is assumed that, the power supply for each mote is a constant 3V; when the node is sending messages, the CPU is in the Active state, LED lights up and radio transmission power is averagely at +4dBm thus it consumes 12mA; when the node is listening for messages, the CPU is in the Active state; when the node is receiving messages, the CPU is in the Active state and the radio current is 8mA. When a node goes into sleep, the CPU is in standby state. The initial energy of each node is set to be the same. When the power assumption of one node exceeds the limit, the node is supposed to be disabled. To simplify the model for simulation, we do not consider the CPU cycle power consumption and consider the battery model a linear one. There are two main simulating parameters to be set, that is, the nodes number, N, and the simulating time, T. When the simulating starts, the nodes are randomly deployed in a square with density of 100 nodes per km² shown in Figure 2 and WSNSim will generate the sensor data in some random distribution according to different applications. The clock is timing, when a timer is fired, an event will occur and WSNSim will check the task queue to perform related operation. The power consumption is calculated and stored for each operation in each node. When the simulation finishes, the energy used in each node will be recorded into files for further analysis.

Figure 2. 1000-Nodes Random Network

4.2. Performance Evaluation

In our simulation, Bernoulli loss process model is used to simulate link characteristics [20]. Each node periodically transmits packet to the sink. If the sink does not receive any data, it launches a requirement for retransmission. The node number N is set from 100 to 1000 for different window k. Performance metric of energy efficiency and packet success rate are collected.

4.2.1. Energy Efficiency

In WSNs, there are N nodes. Residual energy model e_i in node i is denoted by Equation 1:

$$e_i={\mathit{\text{UI}}}_{i}t$$

(1)

There t is the left work time of node i, so the average residual energy of WSNs Ē can be obtained from Equation 2.

$$\overline{E}=\frac{1}{N}\sum _{i=1}^n{e}_i$$

(2)

For different routing protocols, lager Ē means more energy efficiency after WSNs experience same time under the same conditions.

Figure 3 shows the comparison of average energy consumption over 10000 seconds, where node number N=100, k=9. As a result, it is obvious that LQER can save more energy than MHFR and MCR, especially with time passed. This is because it is not enough to consider only cost and hop count in case the link quality is poor. If the poor link is chosen to deliver the data, loss rate will be high and retransmission will cause extra energy consumption, at the cost of lifetime of WSNs.

Figure 3. Average Energy Consumption of MHFR, LRER and MCR

4.2.2. Scalability

The number of sensor nodes deployed in studying a phenomenon may be up to thousands or more. For some special application, the number may reach an extreme value of millions. The new routing algorithms must be able to work with such number of nodes. So it is very important and necessary to test the scalability of protocols for a larger scale of WSNs.

Figure 4 shows the difference of average energy consumption of MHFR and LQER over time with k=9, node number from 0 to 1000. As the node number increases, the difference value between MHFR and LQER also increases, which indicates that LQER has a good scalability of energy efficiency. Results are similar in Figure 5, which shows the difference value of average energy consumption of MCR and LQER.

Figure 4. Difference of Average Energy Consumption in MHFR and LQER

Figure 5. Difference of Average Energy Consumption in MCR and LQER

4.2.3. Data Delivery Efficiency

The success rate is the ratio of number of successfully received data packets at a sink to the total number of data packets generated by a source. This metric shows how effective the data delivery is. Also, it is one of most important parameters of QoS. In some applications such as target tracking, data delivery efficiency outweighs energy efficiency. Unless enough reliable data is transmitted to the base station, the target can not be well tracked and controlled.

Figure 6 shows the comparison of success rate in LQER, MHFR and MCR with different node number, where k equals to 9. It is clear that the success rate in LQER is higher than that in MHFR and MCR, and when node number increases, the variation is small, which indicates a good scalability of data delivery efficiency. The success rate of MHFR and MCR decline as the node number increases and that of MCR declines more quickly. Particularly, when number of nodes equals to 1000, LQER (95.01%) results in percent of success rate that are more than 21.23% higher than those in MHFR (78.37%), and 86.99% higher than those in MCR (50.81%)

Figure 6. Success Rate of MHFR, LQER and MCR

4.2.4. Impact of Window K

The simulation experiments above have been done with k equal to 9. However, as the value of k reflects how much attention is paid to the historical link quality, different values of k result in different simulation result. Figure 7 shows the success rate of LQER and energy consumption of LQER compared to MHFR with the values of k from 3 to 13, where the node number is 500. It is not that the larger k is, the better the protocol performs. If k is too large, it requires for more storage. From Figure 7, we can see that success rate keeps increasing while k is no larger than 9. After 9, the increase is not so obvious and seems stable. Similar is the energy consumption. It tends not to decline when k is larger than 9. Thus k=9 may be a best choice for link quality estimation.

Figure 7. Success Rate of LQER and Energy Consumption of LQER Compared to MHFR

5. Conclusion

References

1. Ma, Z.; Sun, Y. Research on routing protocol of large wireless sensor network. Computer Engineering and Applications (in chinese) 2004, 165–168. [Google Scholar]
2. Ye, F.; Chen, A.; Lu, S.; Zhang, L. Scalable solution to minimum cost forwarding in large sensor networks. Proceedings of 10th International Conference on Computer Communications and Networks.
3. Wang, X.; Bi, D.; Ding, L.; Wang, S. Agent collaborative target localization and classification in wireless sensor networks. Sensors 2007, 7, 1359–1386. [Google Scholar]
4. Chen, J.; Li, S.; Sun, Y. Novel deployment schemes for mobile sensor networks. Sensors 2007, 7, 2907–2919. [Google Scholar]
5. Chen, J.; Cao, K.; Sun, Y.; Yang, X. a. Continuous drug infusion for diabetes therapy: A closed-loop control system design. EURASIP Journal on Wireless Communications and Networking 2008. to appear. [Google Scholar]
6. Akyildiz, I.; Su, W.; Sankarasubramaniam, Y.; Cayirci, E. Wireless sensor networks: A survey. Computer Networks 2002, 38, 393–422. [Google Scholar]
7. Du, S.; Shi, Z.; W, Z.; Chen, K. Using interacting multiple model particle filter to track airborne targets hidden in blind doppler. Journal of Zhejiang University-Science A 2007, 8(8), 1277–1282. [Google Scholar]
8. Akyildiz, I.; Su, W.; Sankarasubramaniam, Y.; Cayirci, E. A survey on sensor networks. IEEE Communications Magazine 2002, 40(8), 102–116. [Google Scholar]
9. Akkaya, K.; Younis, M. A survey on routing protocols for wireless sensor network. Ad Hoc Networks 2005, 3(3), 325–349. [Google Scholar]
10. Heinzelman, W.; Kulik, J.; Balakrishnan, H. Adaptive protocols for information dissemination in wireless sensor networks. Proceedings of the 6th Annual ACM International Conference on Mobile Computing and Networking 1999, 174–185. [Google Scholar]
11. Intanagonwiwat, C.; Govindan, R.; Estrin, D. Directed diffusion: A scalable and robust communication paradigm for sensor networks. Proceedings of the 6th Annual ACM International Conference on Mobile Computing and Networking.
12. Rodoplu, V.; Ming, T. Minimum energy mobile wireless networks. IEEE Journal of Selected Areas in Communications 1999, 17(8), 1333–1344. [Google Scholar]
13. Xu, Y.; Heidemann, J.; Estrin, D. Geography-informed energy conservation for ad hoc routing. Proceedings of the 7th Annual ACM/IEEE International Conference on Mobile Computing and Networking.
14. Yu, Y.; Estrin, D.; Govindan, R. Geographical and energy aware routing: A recursive data dissemination protocol for wireless sensor networks. In Technical Report UCLA-CSD TR-01-0023, UCLA; Computer Science Department, May 2001. [Google Scholar]
15. Heinzelman, W.; Chandrakasan, A.; Balakrishnan, H. Energy-efficient communication protocol for wireless sensor networks. Proceeding of the Hawaii International Conference System Sciences.
16. Stankovic, J.; Abdelzaher, T.; Lu, C.; Sha, L.; Hou, J. Real-time communication and coordination in embedded sensor networks. Proceedings of The IEEE 2003, 91(7), 1002–1022. [Google Scholar]
17. Sohrabi, K.; Gao, J.; Ailawadhi, V.; Pottie, G. Protocols for self-organization of a wireless sensor network. IEEE Personal Communications 2000, 7(5), 16–27. [Google Scholar]
18. Akkaya, K.; Younis, M. Energy and QoS aware routing in wireless sensor networks. Cluster Computing 2005, 8(2-3), 179–188. [Google Scholar]
19. Chu, M.; Haussecher, H.; Zhao, F. Scalable information-driven sensor querying and routing for ad hoc heterogeneous sensor networks. International Journal of High Performance Computing Applications 2002, 393–313. [Google Scholar]
20. Woo, A.; Tong, T.; Culler, D. Taming the underlying challenges of reliable multihop routing in sensor networks. Proceedings of the 1st International Conference on Embedded Networked Sensor Systems 2003, 14–27. [Google Scholar]
21. Hamdaoui, M.; Ramanathan, P. A dynamic priority assignment technique for streams with (m,k)- firm deadlines. IEEE Transaction on Computers 1995, 44(4), 1443–1451. [Google Scholar]
22. Lin, R. Study of Target-Tracking Oriented Wireleess Sensor Networks. PhD thesis, Zhejiang University, 2005. [Google Scholar]
23. Chen, J.; Lin, R.; Sun, Y. Development of simulation environment WSNSim for wireless sensor networks. Chinese Journal of Sensors and Actuators 2005, 19(2), 1–6. [Google Scholar]