^{1}

^{2}

^{2}

^{*}

This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).

Flooding is the simplest and most effective way to disseminate a packet to all nodes in a wireless sensor network (WSN). However, basic flooding makes all nodes transmit the packet at least once, resulting in the broadcast storm problem in a worst case, and in turn, network resources are severely wasted. Particularly, power is the most valuable resource of WSNs as nodes are powered by batteries, then the waste of energy by the basic flooding lessens the lifetime of WSNs. In order to solve the broadcast storm problem, this paper proposes a dynamic probabilistic flooding that utilizes the neighbor information like the numbers of child and sibling nodes. In general, the more sibling nodes there are, the higher is the probability that a broadcast packet may be sent by one of the sibling nodes. The packet is not retransmitted by itself, though. Meanwhile, if a node has many child nodes its retransmission probability should be high to achieve the high packet delivery ratio. Therefore, these two terms—the numbers of child and sibling nodes—are adopted in the proposed method in order to attain more reliable flooding. The proposed method also adopts the back-off delay scheme to avoid collisions between close neighbors. Simulation results prove that the proposed method outperforms previous flooding methods in respect of the number of duplicate packets and packet delivery ratio.

A wireless sensor network (WSN) consists of spatially distributed sensor nodes that cooperatively monitor physical or environmental conditions. They are used in many industrial and civilian application areas, including structural health monitoring, environment and pollutant monitoring, and healthcare applications.

A WSN is composed of two types of nodes: sink nodes and sensor nodes. While sensor nodes collect surrounding information with sensors, a sink node is in charge of connection between the Internet and sensor nodes. A sink node plays an important role as a gateway, so it has powerful and redundant components for the high reliability. On the other hand, sensor nodes are typically equipped with low-end components in consideration of cost, because generally hundreds of or thousands of sensor nodes are needed for a WSN to provide the secure monitoring function.

Besides the low-end components, sensor nodes generally operate with battery power, thus the power is the most important resource in WSN. This is why WSNs have been studied and commercialized based on ZigBee [

Flooding is the most basic and important method for nodes to exchange network information or deliver routing request (RREQ) messages to their destination. The basic flooding, also called blind flooding [

Many flooding mechanisms have been proposed to address the broadcast storm problem. In common, they try to suppress the rebroadcast of duplicate packets based on some basic network information such as location, retransmission probability, or the number of duplicate packets received by each node [

This paper proposes a method which utilizes the number of neighbor nodes in a different way. The proposed method considers the neighbor node condition unlike previous schemes. The neighbor nodes are divided into three types: parent (upper level), sibling (same level), and child (lower level) nodes. This level information can be acquired during several times of initial basic flooding from a sink node. Intuitively, the more siblings a node has, the higher probability it has that its child nodes receive a broadcast packet although it does not retransmit the packet immediately. Thus if a node has many sibling nodes, its retransmission probability may be decreased. Meanwhile, when a node has many child nodes, it has to retransmit a broadcast packet with a higher probability because all the child nodes are highly unlikely covered by the sibling nodes’ retransmission. In short, the retransmission probability in the proposed method is proportional to the number of child nodes and inversely proportional to the number of sibling nodes.

The performance is evaluated using the QualNet 4.5 simulator [

The rest of this paper is organized as follow: Section 2 introduces related works, and Section 3 describes the proposed method using the number of sibling and child nodes. In Section 4, performance of the proposed method is compared with other methods through simulation. Lastly, Section 5 concludes this paper with a discussion of future work.

As mentioned before, the fundamental broadcasting method is basic flooding. It is easy to implement basic flooding, but in a network with

The flooding methods can be divided into two types as shown in

The heuristic-based methods can be further divided into probability-based and area-based ones according to whether location information is adopted or not. The probability-based scheme either fixes the retransmission probability at each node or adjusts it to dynamic network condition, whereas a node in the basic flooding always retransmits a packet if it has not been received before. In the area-based method, a node computes the area it can newly cover with its retransmission, based on the location information of the neighbors and itself. The retransmission is performed only when the newly covered area is larger than any threshold. These heuristic-based methods have shortcomings such as the inflexibility of the retransmission probability or the requirement of location information.

In the mean time, neighbor-information-based, source-tree-based, and cluster-based algorithms belong to the topology-based schemes [

The proposed algorithm is a hybrid of the probability-based method and the neighbor-information-based method. Basically, the proposed method sets the retransmission probability of broadcast packets like the probability-based method, but the probability can be different for each sensor node depending on the neighbor node information. However, the neighbor information is collected just once at the network initial time because sensor nodes are usually assumed to be static. This is the main difference from most topology-based algorithms which have been designed for mobile

In the existing dynamic probabilistic flooding, the retransmission probability is adjusted according to the number of duplicate packets received within a period of time. On the other hand, the proposed method utilizes the neighbor node information to determine the retransmission probability. This neighbor information, however, is more detailed than the previous neighbor-information-based methods. In the previous schemes, each node has just the neighbor node list to check whether all its neighbors have received broadcast packets already. If any neighbors have not received a broadcast packet, the packet is retransmitted. On the other hand, the proposed method classifies neighbor nodes into three classes: parents, siblings, and child nodes.

Intuitively, the more siblings a node has, the less necessity of retransmission it has such that all the children may receive a broadcast packet. Although the node may not retransmit the packet, its children will likely receive the packet from aunt nodes (siblings of the parent). Meanwhile, if a node has many child nodes its retransmission probability should be high to achieve a high packet delivery ratio. Therefore, the two terms—the numbers of siblings and child—are utilized in the proposed method in order to attain more efficient and reliable flooding.

The proposed algorithm is composed of three steps. First, nodes obtain neighbor information through the Hello messages. Second, they determine their level within the topology tree and compute the relation with their all neighbors. Finally, they decide the retransmission probability based on the numbers of child and sibling nodes. After that, each node rebroadcasts a packet according to its retransmission probability.

_{c}_{s}_{p}_{t}_{n}

If a node has no child, the retransmission is not needed. On the other hand, if it has no siblings but just child nodes, it must retransmit a packet. In other situations, the retransmission is performed according to the probability _{t}

The first term, 1/(_{s}

Thus the second term, _{i}N_{c}_{n}_{i}_{i}_{i}_{t}

We discovered some problems of the basic algorithm during the study. In theory, every node should be able to determine its relation with all the neighbors after the first broadcast. To do this, every node must receive the broadcast message and the path from the sink to itself should be permanent. However, this is not true in practical due to frequent collisions and wireless channel bit errors.

As mentioned above, the path from the sink to a node is not always the same. Because of collisions or channel contention, the relation with a neighbor may be changed for every broadcasting packet, e.g., from the parent to a sibling or from a sibling to a child, and so on.

Assuming an ideal sensor deployment and no packet collisions, the relation may not be changed. In practical situations, however, there are so many factors affecting the packet transmission through wireless channels, including collisions and contention, which are more serious in a network with a high node density. Therefore, the more relation changes happen in the denser network.

Lastly, too many packet collisions were observed when the proposed algorithm was adopted, and we found out this was because all sibling nodes attempted the retransmission at the same time the minute they had received a packet. This collision problem can be relieved by letting each node wait a random back-off delay before retransmission.

The proposed algorithm is extended with two additional features. First, we suggest the new concept of the probability that a neighbor may be a parent, a sibling, or a child node during broadcasting: _{parent}_{sibling}_{child}

_{parent}_{parent}_{sibling}_{parent}_{sibling}_{child}_{p}_{s}_{c}_{p}_{s}_{c}

The initial five times of flooding increase energy consumption, compared to the previous suggestion requiring just one initial flooding. However, this may not be too large an overhead, considering that the effect of accurate probabilities on the performance lasts for a long time because the WSN topology is static in usual. In mobile

The second notable feature is the use of the back-off delay before retransmission. The reduction of collisions by the back-off time was observed in _{s}

The proposed algorithm is evaluated using the QualNet 4.5 simulator [^{2} area and the transmission range is 50 m. The IEEE 802.11b MAC is adopted, and the two ray propagation model is used because some sensor signals can be reflected off the ground while some signals follow the LOS (line-of-sight) path. The node mobility is not considered according to the basic concept of the WSN. As the proposed algorithm does not use location information, its performance is compared with only non-location-based methods such as the basic flooding, the fixed probabilistic flooding, and the dynamic probabilistic flooding. The number of nodes changes from 50 to 100 in steps of 10, and each result is the average over 20 experiments. The specific simulation parameters are given in

In the figure, the basic flooding achieves 100% PDR in most cases. However, this is acquired at the cost of node energy because all nodes in this flooding should retransmit the packet at least once. Meanwhile, the PDR of the fixed probabilistic flooding is much different depending on the number of nodes as the retransmission probability is fixed at 0.6 all the time. On the other hand, the dynamic method adjusts its retransmission probability to the node density, so it can achieve PDR greater than 90% irrespective of the node density. The proposed algorithm has the better PDR than the other probabilistic methods, being greater than 95% in all cases. It can more adaptively determine the retransmission probability compared with the others.

With less than 60 nodes, the fixed probabilistic flooding with the retransmission probability of 0.6 generates the least number of broadcast packets. However, this is achieved at the cost of packet delivery ratio. From

From

The basic flooding and the fixed probabilistic flooding complete their work faster than the others as they do not adopt the back-off mechanism. On the other hand, the dynamic probabilistic flooding and the proposed method wait for some time before retransmission. Particularly, the proposed method needs more time than the dynamic probabilistic flooding because the back-off time is defined as _{s}

To sum up, there is a tradeoff between the flooding latency and the PDR. The longer the back-off time is, the less the number of collisions is. Considering that energy is the most important resource in WSN, the proposed algorithm is a proper method in WSN although it takes a little larger flooding latency.

In this paper we have proposed a novel flooding algorithm that can effectively reduce the number of broadcast packets and collision. The proposed method is one of dynamic probabilistic schemes using the numbers of child and sibling nodes. The more child nodes and the less sibling nodes a node has, the higher retransmission probability it has. The proposed method needs initial overhead due to the first five instances of flooding, but it outperformed all the other methods in the aspects of packet delivery ratio and the number of retransmissions. It is worth noticing that the proposed algorithm achieves stable coverage irrespective of the node density.

Some researchers have suggested off-line methods to find an optimal routing tree for packet broadcast, but it is proved as an NP problem [

Implementing the proposed algorithm on a real testbed, we will investigate what modifications are needed for the better performance in a practical network. Particularly, we have to study how to decrease the broadcasting latency without damaging other performance metrics.

This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (Grant No. 2010-0006048).

Power consumption of a MicaZ node [

Taxonomy of flooding methods.

Operation example of the proposed algorithm.

Parent nodes sharing a child.

Ratio of obtained relation information after the first basic flooding.

Example of topology change.

Relation changes between nodes.

Ratio of node relation changes on sequential flooding.

Effect of the back-off delay on the number of received packets.

Example of computing the neighbor node relation.

Packet delivery ratios.

Number of duplicate packets.

Flooding completion time.

Initial probability (_{i}

0 | ∼ | 3 | 1.0 |

4 | ∼ | 5 | 0.9 |

6 | ∼ | 7 | 0.8 |

8 | ∼ | 13 | 0.7 |

14 | ∼ | 30 | 0.6 |

31 | ∼ | 0.5 |

Simulation parameters.

| |
---|---|

Simulator | QualNet 4.5 |

Network range | 250 m × 250 m |

Transmission range | 50 m |

Number of nodes | 50∼100 (steps of 10) |

Bandwidth | 2 Mbps |

Traffic type | CBR |

Packet rate | 1 packet/10 s |

Packet size | 64 bytes |

Simulation time | 300 (S) |

Number of trials | 20 |

Propagation model | Two ray model |

Mobility | None |

MAC Protocol | IEEE 802.11 |

Theoretical node density (nodes/range)

| ||
---|---|---|

50 | 6.40 | 6.35 |

60 | 7.68 | 7.58 |

70 | 8.96 | 9.43 |

80 | 10.24 | 10.58 |

90 | 11.52 | 11.56 |

100 | 12.80 | 13.69 |

| ||

- Network size: 250 m × 250 m | ||

- Transmission range: 50 m |