An efficient tree based self-organizing protocol for internet of things

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. Seehttp://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2578298, IEEE Access

1

An Efficient Tree-based Self-Organizing Protocol forInternet of Things

Tie Qiu, Senior Member, IEEE, Xize Liu, Lin Feng∗, Yu Zhou, Kaiyu Zheng, Student Member, IEEE,

Abstract—Tree networks are widely applied in Sensor Networks of Internet of Things (IoTs). This paper proposes an EfficientTree-based Self-organizing Protocol (ETSP) for sensor networks of IoTs. In ETSP, all nodes are divided into two kinds: network nodesand non-network nodes. Network nodes can broadcast packets to their neighboring nodes. Non-network nodes collect the broadcastedpackets and determine whether to join the network. During the self-organizing process, we use different metrics such as number ofchild nodes, hop, communication distance and residual energy to reach available sink nodes’ weight, the node with max weight will beselected as sink node. Non-network nodes can be turned into network nodes when they join the network successfully. Then a tree-basednetwork can be obtained one layer by one layer. The topology is adjusted dynamically to balance energy consumption and prolongnetwork lifetime. We conduct experiments with NS2 to evaluate ETSP. Simulation results show that our proposed protocol canconstruct a reliable tree-based network quickly. With the network scale increasing, the self-organization time, average hop and packetloss ratio won’t increase more. Furthermore, the success rate of packet in ETSP is much higher compared with AODV and DSDV.

Index Terms—Internet of Things, Self-organization, Tree-based Sensor Networks, Lifetime.

F

1 Introduction

Internet of Things (IoTs) [1], [2] enables objects to collector exchange data using many network technologies, such

as sensor networks, wireless communication, data collection[3], [4], [5], etc. Among them, sensor network is indispensableto IoTs. It has been widely used in localization [6], industrialautomation, environmental monitoring [7] and other applica-tions. Sensor networks consist of a lot of low-cost, low-powertiny sensor nodes which are randomly distributed. Thesenodes can communicate with each other to collect and forwardsensing data. With the scale increasing and devices updating,the network system becomes more and more complex. Thememory, energy and ability of computing are limited bynetwork nodes [8], [9]. In order to maximize lifetime, manyresearchers [10] apply themselves to control network topology[11], build better data transmission route and balance energyconsumption of nodes [12], [13], [14].

Tree network is essentially a combination of bus networkand star network, which can prolong the lifetime of network.Therefore, how to build a tree-based network with a maximumlifetime for sensor networks of IoTs has become a critical issueat present. But choosing a real maximum lifetime tree fromall extended trees is a NP-complete problem [15]. So in orderto meet the requirement of real-time, we need to choose asub-optimal network. In [16], Zhu et al. have proved thata tree-based network cannot be built within a polynomialtime. They construct a spanning tree in polynomial timethrough subset division. Even in the worst case the tree can beconstructed within an exponential time. WSTDO (WeightedSpanning Tree Distributed Optimization) [17] is a distributed

Tie Qiu, Xize Liu, Kaiyu Zheng are with the School ofSoftware, Dalian University of Technology, Dalian, Liaoning,China, 116620 (e-mail: [email protected]; [email protected];[email protected])Lin Feng, Yu Zhou are with the School of Innovation and En-trepreneurship, Dalian University of Technology, Dalian, Liaoning,China, 116024 (e-mail: [email protected]; [email protected])Corresponding author: Lin Feng

data transmission technology based on spanning tree andthe network performance depends on density of nodes. Itachieves a better performance in sparse networks. Ye et al.in [18] have verified that without data aggregation the upperlimit of all one-hop nodes’ energy consumption is 98%. LBT(Load-Balanced and energy-efficient Tree) can maximize thenetwork lifetime. Authors take load-balancing and energy-efficient of one-hop nodes into account to construct the tree-based network. Algorithm LBT can preserve that the energyconsumption of the tree-based network is close to the upperlimit, approximately. Data aggregation technology isn’t usedin above literatures. So these methods increase the energyconsumption and network load when data aggregation occurs.In this paper we use the data aggregation technology in tree-based network to reduce the energy consumption and networkload.

In this paper, an Efficient Self-organization Protocol (ET-SP) in tree-based network is proposed. The network nodes(the nodes that have joined the network) are classified intothree types: root node, sink node, sensor node. In the begin-ning of ETSP, there is only a root node whose hop is zero.Then, the root node searches child nodes by broadcastingpackets. After receiving the broadcast packets, the neigh-boring non-network nodes record the topology informationand use different metrics such as number of child nodes,hop, communication distance and residual energy to reachavailable sink nodes’ weight. Next, the node with max weightis selected as sink node. When non-network nodes join thenetwork successfully, they can be turned into network nodes atonce. Our proposed algorithm can build a tree-based networkquickly. In addition, we adjust the topology dynamically andremove the farthest child node to balance energy consumptionand prolong the whole network lifetime.

The rest of this paper is organized as follows: In Section 2,we discuss the related work and research problem. The energy-efficient self-organization strategy for tree-based networks is

www.redpel.com +917620593389




2

presented in Section 3. Section 4 gives the implementationof ETSP. The experiments and experimental results are dis-cussed in Section 5. Section 6 is the conclusion of this paper.

2 RELATED WORK AND PROBLEM STATE-MENT2.1 Related WorkThe strategies based on topology control can be dividedinto the three types: Multi-node transmit [19], Connecteddominating set [20], Clustering algorithm [21]. Among them,clustering algorithm is widely used. LEACH (Low EnergyAdaptive Clustering Hierarchy) [22] is one of clustering al-gorithms, which creates and maintains clusters to lower theenergy of network. Each node uses a stochastic algorithm todetermine whether it becomes a cluster head. The node withthe maximum energy is selected as the cluster head. HEED(Hybrid Energy-Efficient Distributed Clustering) [23] is alsobased on clustering topology. Except for residual energy,HEED considers the number of neighboring nodes and degreesin cluster head selection. The network is stabile when there areno hot nodes during a period. However, LEACH and HEEDstill reelect the cluster-header after a period, which wastessome energy. In [24], Chen et al. propose an improved LEACH.In this algorithm, it consumes more energy consumption thanLEACH when there is only one hop between cluster andBS (Base Station). A new algorithm named EEDC (Energy-Efficient Distance-based Clustering) is proposed in [25]. FirstEEDC builds a cluster-header candidate set based on theresidual energy of nodes and then selects a best cluster-headerfrom the candidate set based on distance. The simulation re-sults show that EEDC outperforms than LEACH and HEED.Different form HEED, new cluster head is reelected whenold cluster head needs to balance energy consumption inECBDA (Energy efficient Cluster Based Data Aggregation)[26]. The selection of cluster-header isn’t periodic. In [27],Jin et al propose a new algorithm to build data transmissionroutes with multi-path disjoint protocol. The new algorithmimproves the energy-efficiency of nodes and ensures that thenetwork has a higher QoS (Quality of Service). But eachnode contains multi-path that increases the complexity ofthe network management. We extend Energy-efficient Self-organization Routing Strategy (ESRS) [28] for tree-basedwireless sensor networks which is proposed in our previouswork and address the above problems to construct a reliabletree-based network quickly.

2.2 Problem StatementAt present, there are three kinds of route algorithm of Ad hocnetwork:

(i) Table driven routing algorithm [29]: DSDV (Destina-tion Sequenced Distance Vector Routing) [30], WRP(Wireless Routing Protocol);

(ii) Demand driven routing algorithm: AODV (Ad hocOn Demand Distance Vector Routing) [31], DSR (Dy-namic Source Routing), ABR (Associativity BasedRouting), SSA (Signal Stability Based AdaptiveRouting), LBR (Link Life Based Routing);

(iii) Layer type of area routing algorithm: CGSR (Cluster-head Gateway Switch Routing) [32], ZRP (ZoneRouting Protocol).

Node0

Node5

Node6

Node1

Node2

Node4 Node3

Fig. 1: A sensor network topology

For AODV and DSDV, the success rate of packet declinessignificantly with the number of nodes increasing, thereforethey are not suitable for large-scale sensor networks and thenetwork performance will decrease rapidly with the increase ofnode number. Hence, we need a more reliable network modeland route algorithm to improve the reliability of communica-tion in large-scale network.

Eq. 1 is an energy model [33]. Assuming that the distancebetween nodei and nodej is d and the packet length is Lbits, the energy cost of sending L bits data is Ei,j(L, d) andEr,x(L, d) is the energy cost of receiving L bits data. Ec isthe basic energy consumption of send-receive link. dcr is thethreshold of communication distance. e1 and e2 are energyunits, corresponding to d < dcr and d > dcr. The result of thisenergy model is determined by d in practical application. Sowe can select the closest node as sink node, which is good forreducing the energy consumption.

Ei,j (L, d) = Et,x (L, d) + Er,x (L, d) = L (2Ec + eds)

e =

{e1, S = 2 · · · if d < dcr

e2, S = 4 · · · if d > dcr

(1)

In addition, we also consider the average hop of networkwhen selecting sink node. Figure 1 is a sensor network topolo-gy. Node6 can communicate with Node5 and Node1. Node6 iscloser to Node5. If Node6 selects Node5 as sink node its hop is(HNode1 + 2), else if Node6 selects Node1 as sink node its hopis (HNode1 + 1). We assume that d1 = 6m, d2 = 7m, d = 12m,according to Eq. 1 Node6 selects Node5 as sink node is moreenergy-efficient. If d1 = 6m, d2 = 7m, d = 8m, Node6 selectsNode1 as sink node is more energy-saving. So we need to takedistance and hop of nodes into account to select sink node.

The sink node selection also needs to take the number ofchild nodes into account. In Figure 1 we assume that Node6’ssink node can be Node1, Node4 or Node5, but Node1 is thebest. Node1 has 4 child nodes and if Node6 selects Node1 thatwill increase the energy consumption of Node1 and shorten thelifetime of network. If Node6 selects Node5 or Node4 as sinknode that can reduce the energy consumption and mitigatepacket processing pressure of Node1. Here, we assume that





3

residual energy of Node5 is more than Node4 and Node5 isbetter than Node4. For prolonging lifetime we choose Node5as sink node. After a period of time Node6 can reelect Node4as sink node to balance energy consumption of Node5 andNode4.

During the self-organizing process of tree-based networkwe need to take distance, hop, number of child nodes andresidual energy into account. But how to balance these factorsand build a better tree-based network are research points ofthis paper.

3 Efficient Tree-based Self-Organizing Protocol3.1 Network Self-organizationDuring the process of network construction, network nodessearch child nodes through sending broadcast packets. Non-network nodes select sink node according to the receivedbroadcast packets. The selection of sink node balances thehop, residual energy, number of child nodes and distancebetween two nodes. The more hops there are, the more trans-mission times will be. With the increasing of transmissiontimes the total energy consumption will increase. So we needto control the hop of network. If a sink node has more childnodes it will receive more packets and consume more energyduring a certain period. So for balancing energy consumptionwe need to take number of child nodes into account to selectsink node. According to Eq. 1, the distance between twonodes is considered and the energy is lower with the distancedecreasing. We consider Eq. 5 and Eq. 6 in [34] and balancinghop, residual energy, number of child nodes and distancebetween two nodes, and sink nodei’s weight is given by Eq.(2).

Wi = α

Di+ β

Ni + 1+ λEi + δ

Hi + 1(2)

Here, Wi is nodei’s weight. Ei is nodei’s residual energy.Di is the distance between current node and nodei. Ni is thenumber of nodei’s child nodes. We define the hop of root nodeis 0 and other nodes’ hops are their sink nodes’ hops plus 1. Hi

is the hop of nodei. α, β, λ and δ are normalized parametersof these four factors and they are defined as follows: if themaximum transmission distance is 15 m we set α = 15. If themaximum number of child nodes is 5 we set β = 6. The initialenergy of node is random and if the maximum initial energy is20 J we set λ = 1/20. If the maximum hop is 10 we set δ = 11.

Nodes of network can be divided into three types: rootnode, sink node and sensor node. Root node is a special nodewhose energy is unlimited and it is active all the time. Rootnode is the first network node and at the beginning root nodesends broadcast packets to search child nodes. Non-networknode saves the broadcast packets and calculates the weight ofsink nodes based on Eq. 2. Finally, non-network node selectsthe best sink node with maximum weight to join the network.If the best sink node refuses the non-network node to jointhe network, the non-network node needs to select the sub-optimal sink node, third-optimal sink node until it has joinedthe network successfully. If the non-network node cannot jointhe network after scanning all available sink nodes, it needsto clear all available sink nodes’ information and then savesother broadcast packets to reelect an available sink node. Non-network node begins to select child nodes after joining the

network. Centering on the root node we can construct a tree-based network quickly through broadcast searching and thetree-based network balances the hop, residual energy, numberof child nodes and distance of nodes.

3.2 Dynamically adjust topologyIn the following two cases we have to reconstruct the networkpartially.Case 1. Energy consumption.The sink node not only gathers the data of its own sensorbut also aggregates data of all its child nodes, so the energyconsumption is quicker than sensor nodes. The farthest nodewill be deleted when the energy of the sink node drops belowR%. R% is based on the residual energy of last topologychanging, which means the sink node adds or removes a childnode. Removing a child node equals sending packets to informthe child nodes to reelect sink node and at the same timedelete the information of the child nodes from child nodetable. It is benefit to balance the energy consumption if thefarthest child node joins in other branches of the network. Inorder to compute using Eq. 3 we have to know the number ofchild nodes N . A simple example: we assume N = 5 at themoment t0 and residual energy is E0. After a certain time att1 the residual energy is E1 and E1 = 5E0/6. For balancingenergy consumption, we delete the farthest child node. Herewe assume that the farthest child node joins in other branchesof the network and the number of child nodes N = 4. Atmoment t2, the residual energy is E2 and E2 = 4E1/5, weneed to adjust the topology again for energy balance.

R = N

N + 1(3)

There are 7 nodes as shown in Figure 1. In the beginningNode1 is the sink node of Node2, Node5 and Node6. Aftera period of time the residual energy of Node1 reduces onefourth and based on Eq. 3 we delete the farthest child node.Here we assume Node6 is the farthest child node. Node1 sendsa packet to inform Node6 to reelect sink node. Node6 searchessink node through broadcast. After broadcast searching, N-ode6 receives the information of Node4, Node5 and Node1.According to Eq. 2, Node6 calculates the weight of followingthree types of sink nodes.

(i) The weight of Node1 is the largest. Node6 keeps thestate and other nodes do nothing.

(ii) The weight of Node4 or Node5 is the largest. FirstNode6 sends a packet to Node1 to leave the network,and then sends packets to Node4 or Node5 to requestjoining the network.

(iii) The weights of Node1 and Node5 are same or theweights of Node1 and Node4 are same, we needto compare the hop, ratio of residual energy, num-ber of child nodes and distance. The priority s-election standards are: less hop, greater ratio of((residualenergy)/(numberofchildnode + 1)), lessdistance. So Node1 is better. Node1 is its current sinknode so we need not do anything.

(iv) The weights of Node4 and Node5 are same, we needto compare the hop, residual energy with child nodesnumber and distance. Here, we assume all factors ofNode4 and Node5 are same, next we need to check





4

their sink nodes. The sink node of Node5 is Node1 andNode1 is the sink node of Node6. Node6 reelects sinknode is to balance the energy consumption of Node1.If Node6 reelects Node5 as its sink node, the lengthof the packet from Node5 to Node1 will increase.Furthermore, it increases the energy consumptionof Node1 according to Eq. 1. However, Node4 isn’tin the branch of Node1 and if Node6 joins in thebranch of Node4 that can help to balance the energyconsumption of Node1.

In Figure 1, we assume that Node6 selects Node5 as itssink node. After a period of time, Node1 cannot work as a sinknode. Node5 needs to reelect a sink node and within one hoprange there are no other nodes except for child nodes. Node5broadcasts to inform all child nodes to reelect sink node.All child nodes can get the biggest weight of their availablesink nodes based on Eq. 2 and then send the biggest weightto Node5. Node5 selects the biggest value from the receivedweights of all child nodes and informs the node to reelect sinknode. Here we assume that Node6 is the selected node andNode6 reelects Node4 as its sink node. Node5 removes Node6from its child node table and requests joining in the branchof Node6. During the process of network reorganization, if thehop of the sink node is changed it needs to inform all its childnodes to update their hops.

Case 2. Link failure.Child node sends data packets to its sink node periodically

and sink node also periodically sends response packets to itschild nodes to ensure the links are connected. If a sink nodehas not received any data packet from a child node in a certainperiod it judges the link is unsuccessful and removes the childnode from its child node table. If a child node has not receivedany response packet from its sink node in a certain period itwill judge the link is unsuccessful and re-select sink node.

3.3 Network Performance Evaluate

The nodes of sensor networks are deployed randomly in a testarea. All nodes have to construct a network quickly so the self-organizing efficiency is a very important factor [35]. We needto ensure the network is robust and the real-time performanceof data transmission is high when sensor networks are collect-ing data. With the number of hops increasing, the forwardingtime increase. Thus, the average hop of network and energyof nodes are important aspects to evaluate performance for anetwork. The network lifetime is divided into three types [36]:

(i) FND (First Node Dies): The time from network startsto the first node dies.

(ii) LND (Last Node Dies): The time from network startsto the last node dies.

(iii) PND (Percent Nodes Die): The time from networkstarts to percent nodes die, for example, P% nodesdie.

The initial values of energy are different for each node.In this paper, we select PND and define the network isunavailable when some key nodes of the network died.

Algorithm 1 Select the best sink node

1: i← 0, max weight← 0, sink index← 02: while i < ava sink num do3: calculate the weight W4: optional sink[i].weight←W5: i + +6: if max weight < W then7: sink index← i8: max weight←W9: else if max weight = W then

10: if optional sink[i] is greater then11: sink index← i12: max weight←W13: end if14: end if15: end while16: Output:sink index

4 Algorithm Design4.1 Network Self-organization4.1.1 Select the best sink nodeIn the beginning of ETSP, many non-network nodes that arein sleep mode exit in the network. Then, one of the non-network nodes builds a network and turns into a networknode. At the same time, the network node sends some broad-cast packets and switches into network monitor state. Next,non-network nodes which receive the broadcast packet start atimer and save the node ID of network node which sends thebroadcast packet in array optional sink[]. The best sink nodein array optional sink[] is selected based on Eq. 2 until thetimer expires. The detail algorithm is shown in Algorithm 3.Variable max weight is used to record the maximum weightof all available sink nodes. Variable sink index representsthe index of sink node in array optional sink[]. They areinitialize to 0 in Line 1. Variable ava sink num recordsthe number of available sink nodes. The node calculates theweight of all available sink nodes base on Eq. 2 and recordsthe results in optional sink[].weight (Lines 3-4). The nodegets the index of the best sink node by comparing weights(Lines 6-8). If more than one nodes have the max weight atthe same time, it elects the better one by the standard of lesshops, greater ratio of R and less distance (Lines 9-12). Finally,it outputs the results (Line 16).

There is one loop in Algorithm 1 to scan arrayoptional sink[] and the array length is ava sink num.ava sink num is determined by the limited density of net-works, therefore the ava sink num is numerable and thecomplexity of Algorithm 1 is O(n). All nodes can select a bestsink node rapidly.

4.1.2 Non-network nodes join in networkAfter selecting the best sink node, non-network nodes re-quest to join the network. The related process is shown inAlgorithm 2. Variable sink index tmp records the indexof sink node in array optional sink[]. Variable Eava is theresidual energy of node. The boolean variable is net nodeidentifies whether the node is a network node. According to





5

the results of Algorithm 1, the non-network node sends aPT JOIN REQUEST (Packet of non-network node re-quests to join the network) packet to the best sink node (Line1). Then the node starts a timer and saves the received packetsduring the timer (Line 2). When the network node receives thePT JOIN REQUEST packet, it checks whether the arraychild[] (An array to record the ID of current child nodes)is full. If the array child[] is full, the network node sendsa PT DENIED (Packet from sink node to deny the non-network node joins the network) packet to the non-networknode where the PT JOIN REQUEST packet is from.Otherwise, the network node will send a PT ACCEPTED(Packet from sink node to accept the non-network node joinsthe network) packet. Finally, the non-network nodes thatreceive the reply of network decide whether or not to jointhe network based on the type of reply packet. When thereceived packet is PT ACCEPTED, it becomes a networknode (Line 4). If a node receives a PT DENIED packet(Line 5), it needs to reselect a sub-optimal sink node (Lines7-20). If max weight is equal to 0 (Line 21), it means thatthere is no available sink node and the non-network node can’tjoin the network. Then it sets ava sink num to 0 and waitsfor other nodes’ searching (Lines 22-23). Otherwise, it updatessink index and goes to Line 1 to rejoin network (Line 26).

There is one loop to scan array optional sink[] and the ar-ray length is ava sink num. ava sink num is numerableand the complexity of Algorithm 2 is O(n).

4.2 Reorganization of Hot Area4.2.1 Checking residual energy of sink nodeFor prolonging the network lifetime and balancing energyconsumption, ETSP needs to check the sink nodes’ resid-ual energy. When a sink node adds or deletes a child n-ode it updates Eorg (The residual energy of last topol-ogy change) with Eava. The farthest node will be delet-ed when the energy of the sink node drops below R%.The related algorithm is shown in Algorithm 3. Variableenergy check timer records the timer value of energy check.Variable ENERGY CHECK TIMER records the initial-ize value of energy check timer. If energy check timerexpires and the current energy drops below R% (Line 1),the sink node sends a PT DELETE packet to the farthestchild node (Line 2) and updates Eorg (Line 3). Otherwise itexists current procedure (Lines 4-6). If a sink node receives aPT DELETE OK packet from the child node (Line 7), itremoves the record of farthest child node (Line 8) and updatesthe number of child nodes (Line 9). If N is equal to 0 (Line11), it becomes a non-network node (Line 12). Otherwise, itresets the energy check timer for next round (Line 14).

There is no loop in Algorithm 3, thus the complexity isO(1). However, after sending a PT DELETE (Packet fromsink node to delete a child node) packet the node starts a timerto wait the reply and it will cost some time. If a node receivesa PT DELETE packet or doesn’t receive any reply packetduring a period, it needs to reelect sink node.

4.2.2 Process of reorganizationThe node deleted by its sink node due to low energy needs torejoin the network. The related process is shown in Algorithm4 and Algorithm 5. After initialization (Line 1), the node

Algorithm 2 Non-network node requests to join in network

1: Send PT JION REQUEST to nodeoptional sink[sink index]

2: Start a timer. The node receives and saves packets duringthe timer

3: if the node receives a PT ACCEPTED packet then4: is net node← true5: else if the node received a PT DENIED packet then6: i← 0, max weight← 0, sink index tmp← 0;7: while i < ava sink num do8: if optional sink[i].weight ≤

optional sink[sink index].weight && i = sink indexthen

9: if optional sink[i].weight > max weightthen

10: sink index tmp← i11: max weight← optional sink[i].weight12: else if optional sink[i].weight =

max weight then13: if optional sink[i] is greater then14: sink index tmp← i15: max weight ←

optional sink[i].weight16: end if17: end if18: end if19: i + +20: end while21: if max weight = 0 then22: ava sink num← 023: return24: else25: sink index← sink index tmp26: go to Line 127: end if28: end if

broadcasts PT SINK SEARCH messages and starts atimer (Line 2). Before the timer expires, if the node receives areply packet ACK from sink node (Line 5), it saves node ID inarray optional sink[] (Line 6) and updates ava sink num(Line 7). After the broadcast searching, if ava sink num isequal to 0 (Line 10), it means that there is no available sinknode within one hop range. If the node is a non-network node,it goes to Line 1 to keep searching (Line 12). Otherwise, itcan select the best sink node from its child nodes by executingAlgorithm 5 (Line 14). If ava sink num is more than 0 (Line16), it carries out the procedures in Algorithms 1 and 2 to jointhe network (Line 17). If the node’s hop is changed after beinga network node, it needs to inform its child nodes to updatetheir hops before existing current procedure (Lines 18-19).

There is one loop in the Algorithm 4. The total costsare the times of sending PT SINK SEARCH (Broadcastpacket of non-network node to search available sink node)packets and it is limited. So the complexity of Algorithm 4 isO(n). But there is a certain time interval between two packets’sending and it consumes some time.





6

Algorithm 3 Check energy consumption of sink node

1: if energy check timer = 0 && (Eava/Eorg) ≤ (R =N/(N + 1)) then

2: send a PT DELETE packet to the farthest childnode

3: Eorg ← Eava

4: else5: return6: end if7: if sink node receives PT DELETE OK then8: remove the farthest child node’s record9: N ← N − 1

10: end if11: if N = 0 then12: is net node← false13: else14: energy check timer ←

ENERGY CHECK TIMER15: end if

Algorithm 4 Reorganizing the hot area

1: ava sink num ← 0, sink index ← −1, i ← 0,is sink node← false

2: Broadcast PT SINK SEARCH and start a timer3:4: Before timer expires5: if the node receives ACK then6: update optional sink[]7: ava sink num++8: end if9:

10: if ava sink num = 0 then11: if is net node = false then12: go to Line 113: else14: execute algorithm 515: end if16: else17: execute Algorithm 1 and 218: if node’s hop is changed then19: inform child nodes to update hop20: end if21: end if

In Algorithm 4, if there is no available sink node withinone hop range except for child nodes, the node ought toinform all its child nodes to reorganize. After that it selectsa best sink node from its child nodes. The detail informa-tion is shown in Algorithm 5. Variable max child weightis the max weight of all child nodes. The weight of childnode is the maximum weight of its available sink n-odes. Variable max child weight index is the index ofmax child weight in array child[]. At first, the sink n-ode needs to clear array child[] (Line 1), then it sends aPT REELECT SINK packet to all the child nodes for

Algorithm 5 Network for the child nodes

1: clear array child[]2: send PT REELECT SINK to the child nodes3: start a timer4: if a node receives PT REELECT SINK then5: ava sink num← 0 and clears optional sink[];6: broadcast PT SINK SEARCH7: calculate the weights of sink nodes;8: send the biggest weight is to the sink node9: end if

10: the sink node selects the maximum weight from child[]11: send PT REJOIN REQUEST to

child[max child weight index]12: if a child node receives PT REJOIN REQUEST

then13: is net node← false14: execute algorithm 215: if is net node = true then16: add the old sink node to its child node table17: send PT REJOIN OK to the old sink node18: else19: is net node← true20: sent PT REJOIN FAILED to the old sink

node21: end if22: end if23: if sink node receives PT REJOIN OK then24: delete child[max child weight index]25: is net node← true26: return27: else if sink node receives PT REJOIN FAILED

then28: i← 0, max child weight← 029: while i ≤ N do30: if child[i].weight ≤

child[max child weight index].weight && i =max child weight index then

31: if child[i].weight > max child weight then32: max child weight index← i33: max child weight← child[i].weight34: else if child[i].weight = max child weight

then35: max child weight index← i36: max child weight← child[i].weight37: end if38: i++39: end if40: end while41: end if42: if max child weight = 0 then43: return44: else45: sent PT REJOIN REQUEST to

child[max child weight index]46: go to Line 12;47: end if





7

reelecting sink node (Line 2). At the same time, it starts atimer (Line 3). If a node receives a PT REELECT SINKpacket from its sink node (Line 4), it sets ava sink numto 0 and clears optional sink[] (Line 5). Then it broadcastsPT SINK SEARCH messages (Line 6). After that, thechild node needs to calculate the weights of all available sinknodes according to Eq. 2 (Line 7) and sends the biggestweight to its sink node (Line 8). The sink node needs torecord the received information in array child[] and selec-t the maximum weight when the timer expires (Line 10).Then it sends a PT REJOIN REQUEST packet to n-ode child[max child weight index] (Line 11). When thechild node receives a PT REJOIN REQUEST packet(Line 12), it needs to become a non-network node and jointhe network by executing Algorithm 2 (Lines 13-14). If thenode succeeds in rejoining the network (Line 15), it addsthe original sink node to its child node table and sends aPT REJOIN OK packet to the original sink node (Lines16-17). Otherwise it becomes a network node and replays aPT REJOIN FAILED packet to the original sink node(Lines 19-20). If the sink node receives a PT REJOIN OKpacket, go to Line 16 (Line 14), it supposes to delete child nodechild[max child weight index] from the child node tableand sets child[max child weight index] as its sink node(Lines 23-24). It becomes a network node and sends packetsto inform its child nodes to update their hops (Line 25). Ifthe sink node receives a PT REJOIN FAILED packetit needs to select a sub-optimal child node (Lines 30-41). Ifmax child weight is equal to 0 (Line 44), it means there isno available child nodes and the reconstruction fails, it existscurrent procedure to reelect available child nodes (Line 45).Otherwise it sends a PT REJOIN REQUEST packet tochild[max child weight index] and goes to Line 12 (Lines47-49). There are two loops in Algorithm 5, it scans the arrayoptional sink[] and the array length is ava sink num atfirst, so the complexity is O(n). The times of second loop isthe number of child nodes. The number of child nodes is lessthan the length of array child[]. In conclusion, the maximumalgorithm complexity in ETSP is O(n), which is similar toAODV and DSDV.

5 Simulation and AnalysisIn order to validate our proposed model, we utilized NS2 tosimulate. The new ETSP protocol has three functions: self-organize tree-based network, balance energy consumption,reelect sink node. TABLE 1 lists the simulation parameters.

Before building a tree-based network we need to set pa-rameters based on TABLE 1. According to TABLE 1, themaximum communication radius is 15 m so α is 15. The max-imum number of child nodes is 10 so β is 11. The maximuminit-energy is 29 J so λ is 1/29. The maximum hop is 10 so δis 11.

The topology of ETSP is given in Fig. 8 according tothe information of self-organization process. In the simula-tion, 100 sensor nodes are randomly deployed in the area of100m*100m. In Fig. 8a, the root rode whose coordinates are(0,0) is in the border of network. While in Fig. 8b, the rootnode is located at the center of the simulation area. It can beseen that the results of self-organization are different due tothe differ of root node’s location, but every node succeeds in

TABLE 1: Simulation Parameters

Parameter ValueCommunication radius 15m

Maximum times of sending P T JOIN REQUEST 3Initial energy range 20J-29J

Maximum times of sending P T DELET E 3Queue length 2

Period of broadcast searching child node message 0.1sSend packet power 0.66w

Period of searching sink node 0.2sReceive packet power 0.395w

Period of requesting join the network 1.5sSleep power 0.035w

Period of sending P T JOIN REQUEST 0.3sMaximum number of child nodes 10Period of sending P T DELET E 0.3s

Maximum hop 10Period of child node sends data packet to sink node 0.8s

Times of searching child node 10Period of energy balance check 8sTimes of searching sink node 5

P: P ND 50%

TABLE 2: Experiment Settings

Number ofNode

Test AreaSize (m*m)

Root Nodein Border

(m,m)

Root Nodein Center

(m,m)50 70*70 (0,0) (35,35)100 100*100 (0,0) (50,50)200 141*141 (0,0) (70.5,70.5)400 200*200 (0,0) (100,100)600 245*245 (0,0) (122.5,122.5)

joining the network. For the topology in Fig. 8b, the averagehop is 4.35, less than 8.02 hops in Fig. 8a. Furthermore, there-organization time is 3.03 s, faster than 4.62 s in Fig. 8a.The average energy consumption is 0.34J , less than 0.36J inFig. 8a. Thus, ETSP can achieves a better performance if theroot node is in the center of network.

In order to verify the efficiency of ETSP, five groupsexperiments based on the different scales are conducted. Theexperiment settings are shown in TABLE 2. In each group,we do five experiments whose results are shown in Fig. 3,Fig. 4, Fig. 5, Fig. 6, and Fig. 7. Hop indicates that theselection of sink node is based on the least hop which betweenthe node and available sink node. Distance indicates that theselection of sink node is based on the least distance betweenthe node and available sink node. Left Energy indicates thatthe selection of sink node is based on the maximum residualenergy of available sink node. Child Number indicates thatthe selection of sink node is based on the number of each childnode’s available sink node. ETSP indicates that the selectionof sink node is based on Eq. 2 and takes the four factors intoaccount.

It can be observed that they are all linearly increasingwith node number increasing in Fig. 3. But the slope of self-organization time in ETSP is smaller. Therefore ETSP canconstruct a reliable tree-based network quickly in the largescale network. This process will be faster if the root nodein the center of network. In Fig. 4, the increasing trends ofaverage hop are similar except for the networks based onDistance and Left Energy. They are lager when the nodenumber is over 500. Thus, they need more transmission timesand energy consumption increases. If the root node is locatedat the center of network, it needs less average hop to construct





8

(a) Root coordinates: (0,0) (b) Root coordinates: (50,50)

Fig. 2: Topology of self-organization.

50 100 150 200 250 300 350 400 450 500 550 6002

4

6

8

10

12

14

16

Number of Node

Sel

f−org

aniz

atio

n T

ime

(s)

Hop

Distance

Left Energy

Child Number

ETSP

(a) Root Node in the border.

50 100 150 200 250 300 350 400 450 500 550 6002

4

6

8

10

12

Number of Node

Sel

f−org

anzi

tion T

ime

(s)

Hop

Distance

Left Energy

Child Number

ETSP

(b) Root Node in the center.

Fig. 3: Relationship between Self-organization Time and Number of Nodes.

50 100 150 200 250 300 350 400 450 500 550 600

6

8

10

12

14

16

18

20

22

Number of Node

Aver

age

Hops

Hop

Distance

Left Energy

Child Number

ETSP


50 100 150 200 250 300 350 400 450 500 550 6002

4

6

8

10

12

14

16

Number of Node

Aver

age

Hops

Hop

Distance

Left Energy

Child Number

ETSP


Fig. 4: Relationship between Average Hop and Number of Nodes.

a network, which speeds up the self-organization process.Network lifetime which illustrated in Fig. 5 is limited

by energy consumption. The network based on Hop has thelongest lifetime due to sending less packets. While ETSP’slifetime is a little shorter than others. We can conclude fromFig. 5a and Fig. 5b that the network lifetime is longer whenthe root node in the center.

It can be seen in Fig. 6, for the network based on Hop, thenumber of send packets is less than others. Thus, it decreasesnode energy consumption and prolongs the network time.While ETSP is similar to the network based on Distance, Left

Energy and Child Number. Obviously, the location of rootnode doesn’t affect the number of packets. After constructinga network the nodes begin to send and receive data, packetloss is inevitable in this process. The performance of successrate of packet is one of the most critical indicators for a routingprotocol. It can be observed that the network based on ChildNumber and Hop are lower than ETSP in Fig. 7. With thenetwork scale increasing, the success rate of packet in ETSPdoesn’t decline significantly and it’s over 92%.

From the simulation results we know that: the networkbased on Hop can get longer lifetime. However its throughput





9

50 100 150 200 250 300 350 400 450 500 550 600

160

180

200

220

240

260

Number of Node

Net

work

Lif

etim

e (s

)

Hop

Distance

Left Energy

Child Number

ETSP


50 100 150 200 250 300 350 400 450 500 550 600140

160

180

200

220

240

260

Number of Node

Net

work

Lif

etim

e (s

)

Hop

Distance

Left Energy

Child Number

ETSP


Fig. 5: Relationship between Network Lifetime and Number of Nodes.

50 100 150 200 250 300 350 400 450 500 550 600

0.5

1

1.5

2

2.5x 10

5

Number of Node

Num

ber

of

Sen

d P

acket

Hop

Distance

Left Energy

Child Number

ETSP


50 100 150 200 250 300 350 400 450 500 550 600

0.5

1

1.5

2

2.5x 10

5

Number of NodeN

um

ber

of

Sen

d P

acket

Hop

Distance

Left Energy

Child Number

ETSP


Fig. 6: Relationship between Number of Send Packet and Number of Nodes.

50 100 150 200 250 300 350 400 450 500 550 60091

92

93

94

95

96

Number of Node

Pro

bab

ilit

y o

f S

ucc

ess

(%)

Hop

Distance

Left Energy

Child Number

ETSP


50 100 150 200 250 300 350 400 450 500 550 60090

91

92

93

94

95

96

97

Number of Node

Pro

bab

ilit

y o

f S

ucc

ess

(%)

Hop

Distance

Left Energy

Child Number

ETSP


Fig. 7: Relationship between Success Rate of Packet and Number of Nodes.

and success rate of packet is lower. The network based onDistance can get larger throughput and higher success rateof packet, but its self-organization time is longer and averagehop is bigger. The network based on Left Energy is worseand its self-organization time is longer and average hop isbigger than the network based on Distance. The networkbased on Child Number sends less packets and its successrate of packet is lower. The network based on ETSP balancedistance, hop, number of child nodes and residual energy. Theresults in Figure 4 reveal that ETSP can construct a tree-based network quickly. With the network scale increasing, theself-organization time, average hop and packet loss ratio won’tincrease repaidly. During the process of simulation experimen-t, the sink nodes are about 50% of all nodes and key nodesare less than 50%. So we set P = 50 is feasible. Comparedwith each other between Fig. 3a and Fig. 3b, Fig. 4a and Fig.

4b respectively, we can see that the network is worse whenthe root node in the border. The network based on Eq. 2 isreasonable. Although the network lifetime is shorter, the self-organization time, network average hop, packet number andsuccess rate of packet are balanced.

In Fig. 8, we evaluate the performance of ETSP, AODVand DSDV with different number of sensor nodes. The rootnode is located in the center of topology. It can be seen thatthe network lifetime of ETSP is longer than DSDV, becauseit periodically checks the residual energy of sink node and re-organize the hot area to achieve energy consumption. What’smore, the success rate of packet in ETSP is further higher thanAODV and DSDV, it keeps stable with the number of sensornodes increasing. Thus, the network constructed by ETSP isreliable.





10

50 100 150 200 250 300 350 400 450 500 550 600100

150

200

250

Number of Nodes

Net

wor

k lif

etim

e (s

)

ETSPAODVDSDV

(a) Network lifetime.

50 100 150 200 250 300 350 400 450 500 550 6000

10

20

30

40

50

60

70

80

90

100

Number of Nodes

Prob

abilit

y of

Suc

cess

(%)

ETSPAODVDSDV

(b) Success rate of packet.

Fig. 8: Performance Comparison of Three Route Algorithms.

6 ConclusionIn this paper, we propose an efficient self-organization pro-tocol named ETSP for sensor networks of IoTs. ETSP savesmore energy and has a longer network lifetime by constructinga tree-based network quickly. We use the weight of nodes,including residual energy, hop, number of child nodes anddistance between the nodes, to determine whether the nodecan be a sink node. Thus the depth of tree is optimized byusing ETSP. During the process of data transmission, thenetwork topology changes dynamically. Each sink node willbe dynamically reselected due to the energy consumption ofsink nodes is faster than other nodes. The simulation resultsshow that ETSP is able to build reliable tree-based networks,reduces the energy consumption and prolongs the lifetime ofsensor networks.

AcknowledgmentThis work is supported by Natural Science Foundation of P.R.China (Grant No. 61202443) and the Fundamental ResearchFunds for the Central Universities (Grant No. DUT16QY27)

References[1] Zhang DQ, Zhao SJ, Yang LT, Chen M, Wang YS, Liu HZ.

NextMe: Localization Using Cellular Traces in Internet of Things.IEEE Transactions on Industrial Informatics 2015; 11(2):302-312.

[2] Ning H, Liu H, Yang LT. Aggregated-proof based hierarchicalauthentication scheme for the Internet of Things. IEEE Trans-actions on Parallel and Distributed Systems 2015; 26(3):657-667.

[3] Atzori L, Iera A, Morabito G. The internet of things: A survey.Computer networks 2010; 54(15):2787-2805.

[4] Turkanovic M, Brumen B, Holbl M. A novel user authenticationand key agreement scheme for heterogeneous ad hoc wirelesssensor networks, based on the internet of things notion. Ad HocNetworks 2014; 20:96-112.

[5] Tsai CW, Lai CF, Vasilakos AV. Future Internet of Things: openissues and challenges. Wireless Networks 2014; 20(8):2201-2217.

[6] Guangjie Han, Jinfang Jiang, Chenyu Zhang, et al. A Surveyon Mobile Anchor Node Assisted Localization in Wireless SensorNetworks. IEEE Communications Surveys & Tutorials 2016; 1-1.

[7] Songtao Lu, Zhengdao Wang, Zhaohui Wang, and Shengli Zhou.Throughput of underwater wireless Ad hoc networks with ran-dom access: a physical layer perspective. IEEE Transactions onWireless Communications 2015; 14(11):6257-6268.

[8] Khan JA, Qureshi HK, Iqbal A. Energy management in wirelesssensor networks: A survey. Computers & Electrical Engineering2015; 41:159-176.

[9] Silva R, Silva JS, Boavida F. Mobility in wireless sensor network-sĺCSurvey and proposal. Computer Communications 2014; 52:1-20.

[10] Han G, Jiang J, Bao N, et al. Routing protocols for underwaterwireless sensor networks[J]. IEEE Communications Magazine2015; 53(11):72-78.

[11] Qiu T, Chi L, Guo W, et al. STETS: A Novel Energy-efficientTime Synchronization Scheme based on Embedded NetworkingDevices. Microprocessors & Microsystems 2015; 39(8):1285-1295.

[12] Songtao Lu, Vitor Nascimento, Jinping Sun, and ZhuangjiWang. Sparsity-aware adaptive link combination approach overdistributed networks. Electronics Letters 2014; 50(18):1285-1287.

[13] Han G, Qian A, Jiang J, et al. A Grid-Based Joint Routing andCharging Algorithm for Industrial Wireless Rechargeable SensorNetworks. Computer Networks 2016.

[14] Qiu T, Luo D, Xia F, et al. A Greedy Model with Small World forImproving the Robustness of Heterogeneous Internet of Thingsl.Computer Networks 2016.

[15] Wu Y, Fahmy S, Shroff NB. On the construction of a maximum-lifetime data gathering tree in sensor networks: NP-completenessand approximation algorithm. IEEE INFOCOM 2008: the 27thInternational Conference on Computer Communications 2008;1013-1021.

[16] Zhu X, Wu X, Chen G. An exact algorithm for maximumlifetime data gathering tree without aggregation in wireless sensornetworks. Wireless Networks 2015; 21(1):281-295.

[17] Carr-Motyckova L, Dryml D. Distributed Energy Efficient DataGathering without Aggregation via Spanning Tree Optimization.12th International Conference on Ad-hoc, Mobile, and WirelessNetworks 2013; 87-98.

[18] Ye JH, Peng K, Wang CG, Wang YK, Zhang L, Ma XQ, JiangHB. Lifetime optimization by load-balanced and energy efficienttree in wireless sensor networks. Mobile Networks and Applica-tions 2013; 18(4):488-499.

[19] Liang O, Sekercioglu YA, Mani Na. Gateway multipoint relays-an MPR-based broadcast algorithm for ad hoc networks. pro-ceedings of 10th IEEE Singapore International Conference onCommunication systems 2006; 1-6.

[20] Rai M, Verma S, Tapaswi S. A power aware minimum connecteddominating set for wireless sensor networks. Journal of networks2009; 4(6):511-519.

[21] Dagdeviren O, Erciyes K, Cokuslu D. A Merging Clustering Al-gorithm for Mobile Ad Hoc Networks. proceedings of InternationalConference on Computational Science and Its Applications 2006;681-690.

[22] Heinzelman WB, Chandrakasan AP, Balakrishnan H. Anapplication-specific protocol architecture for wireless microsensornetworks. IEEE Transactions on Wireless Communications 2002;1(4):660-670.

[23] Younis O, Fahmy S. Distributed Clustering in Ad-hoc SensorNetworks: A Hybrid, Energy-Efficient Approach. Twenty-thirdAnnualJoint Conference of the IEEE Computer and Communi-cations Societies 2004; 629-640.

[24] Chen YL, Wang NC, Shih YN, Lin JS. A Novel Energy-Efficientand Distance-Based Clustering Approach for Wireless SensorNetworks. Wireless Personal Communications, 2014; 75(1):349-368.





11

[25] Afsar MM, Tayarani-N MH. Distributed Clustering in Ad-hocSensor Networks: A Hybrid, Energy-Efficient Approach. SoftComputing in Industrial Applications, Springer InternationalPublishing, 2014; 177-186.

[26] Ranjani SS, Krishnan SR, Thangaraj C, Devi KV. AchievingEnergy Conservation by Cluster Based Data Aggregation in Wire-less Sensor Networks. Wireless Personal Communications, 2013;73(3):731–751.

[27] Jin RC, Gao T, Song JY, Zou JY, Wang LD. Passive cluster-based multipath routing protocol for wireless sensor networks.Wireless networks, 2013; 19(8):1851–1866.

[28] Feng L, Zhou Y, Qiu T. An energy-efficient self-organizationrouting strategy in tree networks. Proceedings of the 8th Interna-tional Conference on Mobile Multimedia Communications 2015;233-236.

[29] Manoj BS, Murthy CSR. On the use of out-of-band signalingin ad hoc wireless networks. Computer Communications, 2003;26(12):1405-1414.

[30] Perkins C E, Bhagwat P. Highly dynamic destination-sequenceddistance-vector routing (DSDV) for mobile computers. ACM SIG-COMM computer communication review, 1994; 24(4): 234-244.

[31] Perkins C, Belding-Royer E, Das S. Ad hoc on-demand distancevector (AODV) routing. 2003.

[32] Chiang CC, Wu HK, Liu WT, Gerla M. Routing in clusteredmultihop, mobile wireless networks with fading channel. proceed-ings of IEEE SICON, 2014; 177-186.

[33] Khamforoosh K. Clustered balanced minimum spanning tree forrouting and energy reduction in wireless sensor networks. IEEESymposium on Wireless Technology and Applications 2011; 56-59.

[34] Zhang H, Chen P, Gong S. Weighted Spanning Tree clusteringrouting algorithm based on LEACH. 2nd International Confer-ence on Future Computer and Communication, 2010; 2223-2227.

[35] Santhi C, Sharmila D. A self-organized location aware energyefficient protocol for wireless sensor networks. Computers & Elec-trical Engineering, 2015; 41:265-274.

[36] Avokh A, Mirjalily G. Dynamic balanced spanning tree (DBST)for data aggregation in wireless sensor networks. 5th InternationalSymposium on Telecommunications, 2010; 391-396.

Tie Qiu received Ph.D and M.Sc. from DalianUniversity of Technology(DUT), in 2005 and2012, respectively. He received B.E. from In-ner Mongolia University of Technology (IMUT),China, 2003. He is currently Associate Profes-sor at School of Software, Dalian University ofTechnology, China. He was a visiting professorat electrical and computer engineering at IowaState University in US (Jan. 2014–Jan.2015). Heserves as an Editorial Board Member of Journalof Advanced Computer Science & Technology

(JACST), a Gest Editor of Computers and Electrical Engineering (Else-vier journal) and Ad Hoc Networks (Elsevier journal), a Program Chairof iThings2016, a TPC member of Industrial IoT2015 and ICSN16, aWorkshop Chair of CISIS13 and ICCMSE15, a Special Session Chair ofWCC 2012 and CSA13, a TPC Member of AIA13, EEC14, EEC15 andEEC16. He has authored/co-authored 6 books, over 50 scientific papersin international journals and conference proceedings. He has contributedto the development of 2 copyrighted software systems and invented 8patents. He is a senior member of China Computer Federation (CCF)and a Senior Member of IEEE.

Xize Liu He is currently an undergraduate stu-dent in School of Software, Dalian Universityof Technology, Dalian, Liaoning, China. He isan outstanding student of DUT and has joinedin several technology innovations. His researchinterests cover embedded system and internet ofthings .

Lin Feng received the BS degree in electronictechnology from Dalian University of Technolo-gy, China, in 1992, the MS degree in power en-gineering from Dalian University of Technology,China, in 1995, and the PhD degree in mechan-ical design and theory from Dalian university ofTechnology, China, in 2004. He is currently aprofessor and doctoral supervisor in the Schoolof Innovation entrepreneurship, Dalian Universityof Technology, China. His research interests in-clude intelligent image processing, robotics, data

mining, and embedded systems.

Yu Zhou Master. Majored in Software Engi-neer and Computer Application Technology ofDalian University of Technology. Obtained thebachelor degree and master degree, respectively.Researched on Embedded System and WirelessSensor Networks. Published three books andthree papers.Jobbing at the SSG(Software andService Group) of Inteląŕs Asia-Pacific ResearchCenter. Working on the development of SGX(Software Guard Extension) and researching onAE (Application Enclave) for Windows and Linux

system.

Kaiyu Zheng received B.E. from Dalian Univer-sity of Technology, China, in 2014. He is MasterStudent in School of Software, Dalian Universityof Technology (DUT), China. He is an excellentgraduate student of DUT and has been award-ed several scholarships in academic excellenceand technology innovation. He participated in”Open Source Hardware and Embedded Com-puting Contest 2012” and won the First Prize.His research interests cover embedded systemand internet of things.



Education

An efficient tree based self-organizing protocol for internet of things