A Transmission Range Reduction Scheme for Reducing ... - IEEE Xplore
A Transmission Range Reduction Scheme for Reducing Power Consumption in Clustered Wireless Sensor Networks Ehssan Sakhaee, Naoki Wakamiya, Masayuki Murata Graduate School of Information Science and Technology Osaka University, Osaka, Japan Email: {sakhaee, wakamiya, murata}@ist.osaka-u.ac.jp Abstract—In this paper we propose a transmission range reduction scheme for a clustered wireless sensor network (WSN), in order to reduce power consumption while maintaining the network connectivity, particularly in scenarios where consecutive reporting of sensing values is inherent in the network. The platform WSN is based on a multi-hop clustering algorithm, where clusterheads send information to the sink via either gateway nodes or other clusterheads of intermediate clusters towards the sink. The main idea of this scheme is that the sink initially gathers the number of nodes or clusters in the network and instructs the reduction in the transmission range of the sensors until this number, hence connectivity is compromised. The proposed scheme does not require geographical information of nodes, and is independent of propagation model and environmental conditions that may cause non-uniform attenuation to radio signals. Furthermore, the scheme is independent of network topology, and can be applied to both uniform and non-uniform distribution of nodes. Simulation results show the effectiveness of the approach in two different clustering schemes, in regards to reducing energy consumption in the network.
primarily based on network traffic load as a parameter for the adjustment. Clustering techniques have further been used to better manage a WSN, by assigning roles to nodes such as clusterhead (CH), gateway (GW) and clustermembers (CM). This results in a hierarchical network, and allows further reduction in energy consumption by allowing high energy nodes to take more power-intensive roles, and lower energy nodes to do simpler less power-intensive tasks. In this paper we propose the Transmission Range Reduction Scheme (TRRS) suitable for multihop clustered WSNs. TRRS aims at adjusting the transmission range based on a minimum required connectivity of the network. The TRRS has the following assumptions: • •
I. I NTRODUCTION
•
In recent years there has been a rapid increase in the deployment of wireless sensor networks (WSNs), and much research has been done in this field. Particularly an important emphasis has been given on saving energy of these small devices which in general have limited power and need to conserve as much energy as possible and operate for as long as possible. Many protocols have thus been developed which are “energy-efficient”. A sensor node consumes energy in listening, receiving, transmitting, and processing data. To save power consumption in the idle state, timing and synchronization mechanisms have been proposed which effectively allow sensors to sleep and wake up at the right time to sense, receive, and send sensor data [1][2]. In transmitting and receiving data, aggregation and reduction of data are helpful [3], but this also may put heavier processing requirements which results in increased energy consumption. The transmission power a node uses to transmit data is a major contributor of its energy consumption and hence lifetime. Hence it is often highly beneficial to find the lowest transmission range that can retain network connectivity in a WSN while effectively conserving power in transmission throughout the lifetime of the WSN. Previous attempts at adjusting transmission range in WSNs have been proposed in [4][5][6], although these have been
• •
• • • •
Nodes are capable of adjusting their transmission range dynamically (increasing and decreasing). No GPS required. Independence of underlying radio propagation model, and environmental factors that may cause non-uniform signal attenuation. Independence from network topology. i.e. applicable to both uniform and non-uniform topologies. At the maximum transmission range, each node is at least within range of another node and there is at least one path from any arbitrary node to the sink. The WSN adopts a clustering mechanism to initially organize nodes into clusters. Most nodes in the network can only reach the sink in a multihop manner. Low node failure rate. MAC layer independence.
The aim of TRRS is to reduce the total amount of energy a WSN consumes during consecutive rounds of data gathering, once clusters have been formed. We have used two target clustering platforms for implementing our proposed scheme, namely the Hybrid, Energy-efficient, Distributed Clustering (HEED) [7] and the Data-Energy-Clustering and Routing (DECRO) [8]. TRRS reduces the transmission range to a degree where the network connectivity is maintained throughout the network and data gathered throughout the network can be effectively routed to the sink. The TRRS process is initiated
978-1-4244-4148-8/09/$25.00 ©2009 This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.
by the sink, which initially determines the number of nodes in the network, and gradually orders the reduction in transmission range of nodes in the network. The paper is organized as follows. Section II presents a background overview of the two clustering algorithms for which TRRS will be applied to. Here, other transmission power-saving approaches are also mentioned. Section III introduces the workings and mechanisms of TRRS, and its application to the two clustering methods. Section IV performs simulations of TRRS for the two clustering approaches. II. BACKGROUND A. Clustering in WSNs 1) Hybrid, Energy-Efficient, Distributed Clustering (HEED): HEED was proposed in [7]. HEED consists of three protocol phases: initialize, cluster setup and finalize. The current remaining energy at each node is used to probabilistically determine the clusterhead based on the residual energy of the nodes. Routing in HEED is performed in a multi-hop manner. Clusterheads route their data to the sink via other clusterheads in the network. In original HEED, ad hoc routing protocols such as Directed Diffusion [9] or Dynamic Source Routing [10] are recommended for intercluster routing of data to the sink. 2) Data-Energy-Clustering and Routing (DECRO) Scheme: DECRO was proposed in [8]. The clusterhead election cost for a node i is based on both the data size and residual energy of nodes. In DECRO , nodes set their hopcount-to-sink (HoTS) value, i.e. the number of hops to a sink, which reflects their relative distances to the sink, used for routing data. The HoTS is initiated by a broadcast of a HoTSIM (HoTS Information Message) from the sink, with a HoTS value of 0. Every node receiving this message updates its HoTS value to the minimum of all the HoTS values in the HoTSIMs received plus 1. It will then update the HoTS value of HoTSIM to that of its own HoTS value, and rebroadcast the updated HoTSIM to its own neighbors. The HoTS information is used for determining the shortest path to the sink from each cluster, as it tells each CH the next cluster that leads towards the sink. Hence the CH will choose the corresponding GW which leads to the cluster with a lower HoTS CH to route its data to, where it is eventually received by the sink. In DECRO, CHs collect data from their CMs, aggregate, and forward their data towards the sink. Routing of data is initiated by the clusters which have the highest HoTS CH in the network, termed highest clusters (HCs) followed by intermediate clusters towards the sink, termed lower neighboring clusters (LNCs). This is depicted in Fig. 1. Each clusterhead of a higher neighboring cluster (HNC) sends its data to the lower HoTS clusterheads of LNCs via gateways leading to such clusterheads. If a LNC does not receive data from higher clusters after a period of time, it will initiate sending its data towards the sink.
Fig. 1.
DECRO cluster formation.
is proposed which uses two power levels, “low” for intracluster communication, and “high” for intercluster communication. Furthermore, it implements transmission timing mechanisms to avoid packet collisions and reduce energy consumption. For the purpose of intercluster routing, a traditional flooding technique such as Ad hoc On-Demand Distance Vector (AODV) [12] is often used to route packets to the sink via clusterheads in the network. In [13], a distributed position-based network protocol is presented to obtain the global minimum energy solution for static networks in a self-reconfiguring manner. This scheme however requires the position information for this purpose, requiring a global positioning system (GPS) equipped at each sensor. √ In [14] an optimal transmission range of Rt = (1 + 5)c is proposed for a uniformly distributed topology, where each cell size c by c in the network contains at least one node. This means that at Rt the network is guaranteed to be connected. The cell here is analogous to a cluster size where each node is capable of reaching every other node. The proposed protocol, COMPOW [15] also aims at determining the minimum power level to use across an entire network for complete connectivity. This scheme is applied to a flat network. In [7], the optimal transmission range is also calculated based on a uniform topology, where it is assumed that at least one node exists in a predefined cell size. The benefits of adjusting transmission range of sensor networks for the purpose of saving energy has been further demonstrated in [4][5][6], where network traffic load is the basis for transmission range adjustment. Centralized and distributed transmit power adjustment schemes are proposed in [16] which adjusts transmit power based on a connected topology. The algorithm uses distance between node pairs, for identifying a connected graph, however this assumes free space or uniform environmental clutter in order to correctly discover distances between nodes.
B. Other Transmission Power-Saving Approaches
III. T RANSMISSION R ANGE R EDUCTION S CHEME (TRRS)
There exist recent works which aim at reducing transmission power in wireless sensor networks. In [11], a clustering scheme
The generalized TRRS process has the following outline. This is slightly adjusted depending on the underlying clus-
978-1-4244-4148-8/09/$25.00 ©2009 This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.
tering protocol in which TRRS is being implemented over. Clustering takes place between steps 1 and 2. The flowchart for the TRRS process is shown in Fig. 3. 1) A sink initiates transmission range reduction by broadcasting a HoTS-PL (HoTS-Power Level) message in the basic format < HoT S, powerLevel, modeT RRS > which is propagated similar to that of the HoTSIM in DECRO. This is shown in Fig. 2. The message contains the HoTS field, a field that contains the desired transmission power level in which the sink wishes the nodes to transmit their data at, and the TRRS mode, which specifies whether data gathering should or should not take place during the TRRS process. 2) The WSN sends reports to the sink, where the sink can then calculate the total number of clusters C in case that TRRS is applied to the intercluster range, or the total number of nodes K, in case that TRRS is applied to the clustering range. Optional collection of data can also take place at this stage, if requested by the sink in the initial HoTS-PL message. Depending on the underlying clustering protocol, reclustering may also take place at this stage, if requested by the sink in the initial HoTS-PL message by the specified TRRS mode. 3) The sink then broadcasts consecutive HoTS-PL messages each time decrementing the power level in the HoTS-PL message by one level, while the reduce transmission range (RTR) variable is set to 1, and retrieves reports from the WSN until the number of current clusters c or number of nodes k retrieved via current reports becomes smaller than C or K, which was initially retrieved at maximum transmission range. It will then set the RTR variable to zero. 4) The sink then broadcasts a final HoTS-PL message, incrementing the power level by one more than the previous value, in order to restore network connectivity. Data gathering and reporting proceeds as normal following the final HoTS-PL message for the current topology.
Fig. 2.
A. Application of TRRS to DECRO
Fig. 3.
The application-specific algorithm of TRRS to DECRO is shown in Fig. 4. Here, the sink initially broadcasts a HoTSPL (HoTS-PowerLevel) message with the highest transmission range that sensor nodes in a WSN can apply, and a HoT S0 value of 0. Upon receiving the HoTS-PL, node n sets its HoTS value as the minimum of all HoTS-PL received plus 1, and the current transmission power (powerLeveln ) and TRRS mode (modeT RRS ) according to the powerLevel value and TRRS mode specified in the received HoTS-PL message. Node n rebroadcasts a HoTS-PL message first updating the HoTS field to its own (HoT Sn ). Nodes in a set Sn , i.e. the set of nodes within the range r(powerLeveln ) corresponding to the current powerLeveln of node n set their HoTS values and their current transmission power accordingly. Eventually, all nodes in the WSN set their HoTS value at their relative distance to the sink and the transmission power P owerLeveln at the identical
HoTS-PL broadcast.
Generalized TRRS flowchart.
maximum value. Simultaneously, nodes will begin the clustering process at the transmission power P owerLeveln . Then, the reporting of network status takes place. CMs send a Cluster Member Data Message (CM D), < N ID, CHID, HoT S(CHi...n ), data/no data >, containing the Node ID, Cluster ID, and also they may or may not append data depending on the mode of TRRS. Once a CH obtains CMDs from its CMs, it sends a TRRS report message containing the number of CMDs they received in their cluster, along with its CH identifier (CHID), and depending on the mode of TRRS, the message may also contain the aggregated data from CMs had the CMs provided data in their CMD. The message is forwarded towards the sink via gateways and intermediate CHs as depicted in Fig. 5. TRRS report message from the CH may contain the aggregated data from CMs. In
978-1-4244-4148-8/09/$25.00 ©2009 This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.
Fig. 6. HEED-ER-TRRS: clusterheads send their CHID and optional data to the sink for calculation of total number of clusters. Fig. 4.
TRRS algorithm for DECRO.
sink for the current topology. B. Application of TRRS to HEED
Fig. 5. DECRO-TRRS: Forwarding number of reports in each cluster to the sink for calculation of total number of nodes.
Section IV, we refer to the scheme where data is contained in the CMD during the TRRS process as TRRS(D), and without data as simply TRRS. Since in DECRO, a GW node belonging to multiple clusters sends its CMD to its best CH, i.e. the CH with the lowest HoTS, the number of CMDs obtained by a CH could be different from the number of nodes in the cluster. This is shown in Fig. 5. After some waiting time tw , the sink calculates the total possible number of nodes Tmax at maximum powerLevel. For example, as in Fig. 5, the sink may receive three report messages with 15, 15 and 10 in their corresponding number of reports field at maximum transmission range of nodes. Hence the total number of nodes Tmax in the network would be calculated as 40 by the sink. The sink continues sending HoTS-PL messages with power level one less than the previous, until the current total number of nodes Tc is less than Tmax of 40 nodes. This means that one or more nodes are isolated from a network with the current cluster organization. In that case the transmission power needs to be increased again by one level, by broadcasting the new and final HoTS-PL message by the
Here, TRRS is applied to the HEED algorithm proposed in [7], as it is a well studied protocol and well known in the field of ad hoc and sensor networks. However, in this paper the routing in HEED is modified to enhance its performance comparable with DECRO. In the modified HEED, the node’s HoTS values are first set, as illustrated in Fig. 2. Furthermore, for routing data to the sink, a CH will send its data to the neighboring CH with the lowest HoTS and highest energy, rather than using traditional approaches such as DSR [10] and Directed Diffusion [9] as suggested in the original paper of HEED [7]. This modified version of HEED is termed HEED-Enhanced Routing (HEED-ER). We note that the actual clustering formation of HEED-ER is identical to that of HEED. Unlike the DECRO algorithm, where there is no distinction between cluster transmission range and intercluster transmission range, CHs in HEED/HEED-ER use a higher transmission range for intercluster communication as they are required to send data directly to a neighboring CH. TRRS here is applied to the intercluster communication range in order to find the minimum communication range required for routing to take place. i.e. for network wide connectivity of clusterheads. Hence TRRS applied to HEED-ER follows the following steps: 1) Clustering formation as that performed in original HEED. 2) Sink broadcasts a HoTS-PL message to set the HoTS value of nodes in the network. 3) CMs send their reports to the corresponding CHs and CHs send reports < CHID >< data(optional) > to the sink at maximum intercluster transmission range. Data may be contained in the report message depending on the TRRS mode specified in the HoTS-PL message. 4) Sink obtains the total number of clusters in the network based on the number of reports it obtains. 5) Sink sends HoTS-PL reducing transmission range until the total number of clusters change.
978-1-4244-4148-8/09/$25.00 ©2009 This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.
6) Sink sends HoTS-PL increasing transmission range by one power level to restore original number of clusters. Fig. 6 shows the principle application of TRRS to HEEDER. We note that in contrast to DECRO-TRRS, TRRS application to HEED-ER does not require reclustering while the intercluster transmission range is being adjusted. This is due to the fact that clustering and routing are independent in HEEDER. Similarly, there are two options when performing the routing of data. For step 2, there is a choice of 1) reporting data to the sink simultaneously with finding the best range - TRRS with data or 2) first finding the best intercluster range, followed by data routing to the sink once the lowest transmission power that maintains complete network connectivity has been identified. These two approaches are termed TRRS(D) and TRRS in the simulations of Section IV. We note that the two modes of operation of TRRS have their respective advantage and disadvantage. The first option TRRS(D), eliminates the delay in finding the best transmission power as data are still being transmitted during the TRRS process. However the energy cost will be relatively high as CHs are transmitting data at higher transmission power than necessary. The second option of pure TRRS involves shorter messages from CHs only to discover the number of CHs reachable by the sink whilst the intercluster range is being decreased. CHs will route the messages by the same rules as when data was included in the packets. Although this method introduces the delay before transmission of data to the sink, it may more effectively save the energy of sensor nodes. Generally, if the amount of data to be sent is relatively large and sensor nodes have limited energy source, pure TRRS should be performed without data being sent during the TRRS process, and followed by reporting data at the final determined minimum transmission range. IV. S IMULATIONS The simulation is implemented in Java. An ideal MAC is assumed to exercise a MAC-independent protocol. Nodes initially have random data size and random energies. 4000 nodes are randomly distributed across 800 by 800 unit area. For DECRO and DECRO-TRRS, an initial range of 450 units is set. TRRS finds a best intercluster communication range of 380 units after 14 rounds of reclustering. At each round the cluster range is reduced by 5 units. DECRO proceeds with clustering at a clustering range of 450 units. Data aggregation factor is set to 50%. For HEED-ER, HEED-ER-TRRS and HEED-ER-TRRS(D), a clustering range of 390 units is used. The intercluster communication range is set to 800 units for all the three simulated protocols. A 50% data aggregation factor is used. HoTS messages are 16-bits, clustering Hello messages (e.g. data size, messages, CH claim message, GW claim message) are set to 32-bits, data sizes are between 0 - 8000 bits. HEEDER-TRRS(D) messages are 24 bits, and DECRO-TRRS(D) messages are 16-bits long. Energy consumption uses the model in [17], which uses 100 pJ/bit/unit2 amount of energy a node spends for broadcasts.
Fig. 7.
Fig. 8.
DECRO and DECRO-TRRS comparison.
HEED-ER and HEED-ER-TRRS comparison.
From Fig. 7, for the first 16 reports DECRO and DECROTRRS perform better than DECRO-TRRS(D). Pure DECRO does not expend energy in finding the best transmission range and this is reflected in the first initial reporting. Additionally the delay in finding the best range is omitted in pure DECRO. However the advantages are overshadowed after 16 reports and DECRO, performing at a higher transmission range than necessary uses up more network energy as shown in Fig. 7. DECRO-TRRS(D) which performs reclustering and data reporting whilst finding the best path has degraded performance for the first 16 reports, however since the best path is discovered, it outperforms pure DECRO after 16 reports. DECRO-TRRS(D), although it has the additional reclustering cost initially to find the minimum transmission range, once this transmission range is found, reporting can be done with the minimum required transmission range and power in oppose to pure DECRO which uses the higher transmission range throughout all the reporting. In DECRO-TRRS(D), the initial overhead of discovering the lowest required transmission range does not result in delay as the CHs are still broadcasting their aggregated data to the sink. The total energy includes energy consumed in HoTS initiation, clustering, and reporting (routing) of data to the sink. DECRO-TRRS which finds the lowest transmission range before data reporting outperforms
978-1-4244-4148-8/09/$25.00 ©2009 This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.
both pure DECRO and DECRO-TRRS(D) at the cost of the short delay in finding best transmission range (14 reclustering rounds without any data being sent). However if delay is not an issue in the startup of the system, DECRO-TRRS is the most effective approach in saving energy in the network as time progresses. In Fig. 8, HEED-ER-TRRS outperforms HEED-ER soon after the first report. As with HEED-ER the same intercluster transmission range is used throughout all consecutive reporting. HEED-ER-TRRS(D) does not use any additional overhead in discovering the lowest intercluster communication range, as this is done while data reporting is taking place during the discovery process. HEED-ER-TRRS first finds the minimum required intercluster range before initiating data reporting. Although HEED-ER-TRRS introduces the delay in initiating reporting, it outperforms HEED-ER and HEED-ER-TRRS(D) in regards to total network energy consumption. However it is important to note that the relative performance of TRRS and TRRS(D) is related to the number of rounds it takes to find the lowest required transmission range. Also no reclustering overhead is performed in both versions of HEED-ER-TRRS which shows its strong performance from initial reporting of data. We note that in regards to HEED-ER-TRRS, the messages broadcasted by the clusterheads to the sink are 16-bits, whereas the aggregated data messages sent by the clusterheads are much larger, as data sent to the clusterhead from each member of the cluster is up to 8000 bits and even after aggregation the relative sizes of packets are multifold. V. C ONCLUSION AND F UTURE W ORK In this paper we introduce the transmission range reduction scheme (TRRS) which aims at reducing power consumption in clustered wireless sensor networks. The application of TRRS was applied to two clustering algorithms, namely DECRO and HEED-ER, and its effectiveness was evaluated in regards to energy consumption. In both cases the effectiveness of TRRS was evident particularly when the rounds of data reporting to sink increases in time. We note that although the scheme reduces only the transmission power of nodes, even in the case that energy consumption is higher for reception, energy constraint devices still benefit by reducing the energy consumed in transmission in any communication network. This becomes more apparent in well or semi-synchronized networks, where nodes awake for data reception at the time a node is expected to transmit. Hence the multifold benefit of transmission power reduction can be applied to even low powered devices (
primarily based on network traffic load as a parameter for the adjustment. Clustering techniques have further been used to better manage a WSN, by assigning roles to nodes such as clusterhead (CH), gateway (GW) and clustermembers (CM). This results in a hierarchical network, and allows further reduction in energy consumption by allowing high energy nodes to take more power-intensive roles, and lower energy nodes to do simpler less power-intensive tasks. In this paper we propose the Transmission Range Reduction Scheme (TRRS) suitable for multihop clustered WSNs. TRRS aims at adjusting the transmission range based on a minimum required connectivity of the network. The TRRS has the following assumptions: • •
I. I NTRODUCTION
•
In recent years there has been a rapid increase in the deployment of wireless sensor networks (WSNs), and much research has been done in this field. Particularly an important emphasis has been given on saving energy of these small devices which in general have limited power and need to conserve as much energy as possible and operate for as long as possible. Many protocols have thus been developed which are “energy-efficient”. A sensor node consumes energy in listening, receiving, transmitting, and processing data. To save power consumption in the idle state, timing and synchronization mechanisms have been proposed which effectively allow sensors to sleep and wake up at the right time to sense, receive, and send sensor data [1][2]. In transmitting and receiving data, aggregation and reduction of data are helpful [3], but this also may put heavier processing requirements which results in increased energy consumption. The transmission power a node uses to transmit data is a major contributor of its energy consumption and hence lifetime. Hence it is often highly beneficial to find the lowest transmission range that can retain network connectivity in a WSN while effectively conserving power in transmission throughout the lifetime of the WSN. Previous attempts at adjusting transmission range in WSNs have been proposed in [4][5][6], although these have been
• •
• • • •
Nodes are capable of adjusting their transmission range dynamically (increasing and decreasing). No GPS required. Independence of underlying radio propagation model, and environmental factors that may cause non-uniform signal attenuation. Independence from network topology. i.e. applicable to both uniform and non-uniform topologies. At the maximum transmission range, each node is at least within range of another node and there is at least one path from any arbitrary node to the sink. The WSN adopts a clustering mechanism to initially organize nodes into clusters. Most nodes in the network can only reach the sink in a multihop manner. Low node failure rate. MAC layer independence.
The aim of TRRS is to reduce the total amount of energy a WSN consumes during consecutive rounds of data gathering, once clusters have been formed. We have used two target clustering platforms for implementing our proposed scheme, namely the Hybrid, Energy-efficient, Distributed Clustering (HEED) [7] and the Data-Energy-Clustering and Routing (DECRO) [8]. TRRS reduces the transmission range to a degree where the network connectivity is maintained throughout the network and data gathered throughout the network can be effectively routed to the sink. The TRRS process is initiated
978-1-4244-4148-8/09/$25.00 ©2009 This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.
by the sink, which initially determines the number of nodes in the network, and gradually orders the reduction in transmission range of nodes in the network. The paper is organized as follows. Section II presents a background overview of the two clustering algorithms for which TRRS will be applied to. Here, other transmission power-saving approaches are also mentioned. Section III introduces the workings and mechanisms of TRRS, and its application to the two clustering methods. Section IV performs simulations of TRRS for the two clustering approaches. II. BACKGROUND A. Clustering in WSNs 1) Hybrid, Energy-Efficient, Distributed Clustering (HEED): HEED was proposed in [7]. HEED consists of three protocol phases: initialize, cluster setup and finalize. The current remaining energy at each node is used to probabilistically determine the clusterhead based on the residual energy of the nodes. Routing in HEED is performed in a multi-hop manner. Clusterheads route their data to the sink via other clusterheads in the network. In original HEED, ad hoc routing protocols such as Directed Diffusion [9] or Dynamic Source Routing [10] are recommended for intercluster routing of data to the sink. 2) Data-Energy-Clustering and Routing (DECRO) Scheme: DECRO was proposed in [8]. The clusterhead election cost for a node i is based on both the data size and residual energy of nodes. In DECRO , nodes set their hopcount-to-sink (HoTS) value, i.e. the number of hops to a sink, which reflects their relative distances to the sink, used for routing data. The HoTS is initiated by a broadcast of a HoTSIM (HoTS Information Message) from the sink, with a HoTS value of 0. Every node receiving this message updates its HoTS value to the minimum of all the HoTS values in the HoTSIMs received plus 1. It will then update the HoTS value of HoTSIM to that of its own HoTS value, and rebroadcast the updated HoTSIM to its own neighbors. The HoTS information is used for determining the shortest path to the sink from each cluster, as it tells each CH the next cluster that leads towards the sink. Hence the CH will choose the corresponding GW which leads to the cluster with a lower HoTS CH to route its data to, where it is eventually received by the sink. In DECRO, CHs collect data from their CMs, aggregate, and forward their data towards the sink. Routing of data is initiated by the clusters which have the highest HoTS CH in the network, termed highest clusters (HCs) followed by intermediate clusters towards the sink, termed lower neighboring clusters (LNCs). This is depicted in Fig. 1. Each clusterhead of a higher neighboring cluster (HNC) sends its data to the lower HoTS clusterheads of LNCs via gateways leading to such clusterheads. If a LNC does not receive data from higher clusters after a period of time, it will initiate sending its data towards the sink.
Fig. 1.
DECRO cluster formation.
is proposed which uses two power levels, “low” for intracluster communication, and “high” for intercluster communication. Furthermore, it implements transmission timing mechanisms to avoid packet collisions and reduce energy consumption. For the purpose of intercluster routing, a traditional flooding technique such as Ad hoc On-Demand Distance Vector (AODV) [12] is often used to route packets to the sink via clusterheads in the network. In [13], a distributed position-based network protocol is presented to obtain the global minimum energy solution for static networks in a self-reconfiguring manner. This scheme however requires the position information for this purpose, requiring a global positioning system (GPS) equipped at each sensor. √ In [14] an optimal transmission range of Rt = (1 + 5)c is proposed for a uniformly distributed topology, where each cell size c by c in the network contains at least one node. This means that at Rt the network is guaranteed to be connected. The cell here is analogous to a cluster size where each node is capable of reaching every other node. The proposed protocol, COMPOW [15] also aims at determining the minimum power level to use across an entire network for complete connectivity. This scheme is applied to a flat network. In [7], the optimal transmission range is also calculated based on a uniform topology, where it is assumed that at least one node exists in a predefined cell size. The benefits of adjusting transmission range of sensor networks for the purpose of saving energy has been further demonstrated in [4][5][6], where network traffic load is the basis for transmission range adjustment. Centralized and distributed transmit power adjustment schemes are proposed in [16] which adjusts transmit power based on a connected topology. The algorithm uses distance between node pairs, for identifying a connected graph, however this assumes free space or uniform environmental clutter in order to correctly discover distances between nodes.
B. Other Transmission Power-Saving Approaches
III. T RANSMISSION R ANGE R EDUCTION S CHEME (TRRS)
There exist recent works which aim at reducing transmission power in wireless sensor networks. In [11], a clustering scheme
The generalized TRRS process has the following outline. This is slightly adjusted depending on the underlying clus-
978-1-4244-4148-8/09/$25.00 ©2009 This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.
tering protocol in which TRRS is being implemented over. Clustering takes place between steps 1 and 2. The flowchart for the TRRS process is shown in Fig. 3. 1) A sink initiates transmission range reduction by broadcasting a HoTS-PL (HoTS-Power Level) message in the basic format < HoT S, powerLevel, modeT RRS > which is propagated similar to that of the HoTSIM in DECRO. This is shown in Fig. 2. The message contains the HoTS field, a field that contains the desired transmission power level in which the sink wishes the nodes to transmit their data at, and the TRRS mode, which specifies whether data gathering should or should not take place during the TRRS process. 2) The WSN sends reports to the sink, where the sink can then calculate the total number of clusters C in case that TRRS is applied to the intercluster range, or the total number of nodes K, in case that TRRS is applied to the clustering range. Optional collection of data can also take place at this stage, if requested by the sink in the initial HoTS-PL message. Depending on the underlying clustering protocol, reclustering may also take place at this stage, if requested by the sink in the initial HoTS-PL message by the specified TRRS mode. 3) The sink then broadcasts consecutive HoTS-PL messages each time decrementing the power level in the HoTS-PL message by one level, while the reduce transmission range (RTR) variable is set to 1, and retrieves reports from the WSN until the number of current clusters c or number of nodes k retrieved via current reports becomes smaller than C or K, which was initially retrieved at maximum transmission range. It will then set the RTR variable to zero. 4) The sink then broadcasts a final HoTS-PL message, incrementing the power level by one more than the previous value, in order to restore network connectivity. Data gathering and reporting proceeds as normal following the final HoTS-PL message for the current topology.
Fig. 2.
A. Application of TRRS to DECRO
Fig. 3.
The application-specific algorithm of TRRS to DECRO is shown in Fig. 4. Here, the sink initially broadcasts a HoTSPL (HoTS-PowerLevel) message with the highest transmission range that sensor nodes in a WSN can apply, and a HoT S0 value of 0. Upon receiving the HoTS-PL, node n sets its HoTS value as the minimum of all HoTS-PL received plus 1, and the current transmission power (powerLeveln ) and TRRS mode (modeT RRS ) according to the powerLevel value and TRRS mode specified in the received HoTS-PL message. Node n rebroadcasts a HoTS-PL message first updating the HoTS field to its own (HoT Sn ). Nodes in a set Sn , i.e. the set of nodes within the range r(powerLeveln ) corresponding to the current powerLeveln of node n set their HoTS values and their current transmission power accordingly. Eventually, all nodes in the WSN set their HoTS value at their relative distance to the sink and the transmission power P owerLeveln at the identical
HoTS-PL broadcast.
Generalized TRRS flowchart.
maximum value. Simultaneously, nodes will begin the clustering process at the transmission power P owerLeveln . Then, the reporting of network status takes place. CMs send a Cluster Member Data Message (CM D), < N ID, CHID, HoT S(CHi...n ), data/no data >, containing the Node ID, Cluster ID, and also they may or may not append data depending on the mode of TRRS. Once a CH obtains CMDs from its CMs, it sends a TRRS report message containing the number of CMDs they received in their cluster, along with its CH identifier (CHID), and depending on the mode of TRRS, the message may also contain the aggregated data from CMs had the CMs provided data in their CMD. The message is forwarded towards the sink via gateways and intermediate CHs as depicted in Fig. 5. TRRS report message from the CH may contain the aggregated data from CMs. In
978-1-4244-4148-8/09/$25.00 ©2009 This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.
Fig. 6. HEED-ER-TRRS: clusterheads send their CHID and optional data to the sink for calculation of total number of clusters. Fig. 4.
TRRS algorithm for DECRO.
sink for the current topology. B. Application of TRRS to HEED
Fig. 5. DECRO-TRRS: Forwarding number of reports in each cluster to the sink for calculation of total number of nodes.
Section IV, we refer to the scheme where data is contained in the CMD during the TRRS process as TRRS(D), and without data as simply TRRS. Since in DECRO, a GW node belonging to multiple clusters sends its CMD to its best CH, i.e. the CH with the lowest HoTS, the number of CMDs obtained by a CH could be different from the number of nodes in the cluster. This is shown in Fig. 5. After some waiting time tw , the sink calculates the total possible number of nodes Tmax at maximum powerLevel. For example, as in Fig. 5, the sink may receive three report messages with 15, 15 and 10 in their corresponding number of reports field at maximum transmission range of nodes. Hence the total number of nodes Tmax in the network would be calculated as 40 by the sink. The sink continues sending HoTS-PL messages with power level one less than the previous, until the current total number of nodes Tc is less than Tmax of 40 nodes. This means that one or more nodes are isolated from a network with the current cluster organization. In that case the transmission power needs to be increased again by one level, by broadcasting the new and final HoTS-PL message by the
Here, TRRS is applied to the HEED algorithm proposed in [7], as it is a well studied protocol and well known in the field of ad hoc and sensor networks. However, in this paper the routing in HEED is modified to enhance its performance comparable with DECRO. In the modified HEED, the node’s HoTS values are first set, as illustrated in Fig. 2. Furthermore, for routing data to the sink, a CH will send its data to the neighboring CH with the lowest HoTS and highest energy, rather than using traditional approaches such as DSR [10] and Directed Diffusion [9] as suggested in the original paper of HEED [7]. This modified version of HEED is termed HEED-Enhanced Routing (HEED-ER). We note that the actual clustering formation of HEED-ER is identical to that of HEED. Unlike the DECRO algorithm, where there is no distinction between cluster transmission range and intercluster transmission range, CHs in HEED/HEED-ER use a higher transmission range for intercluster communication as they are required to send data directly to a neighboring CH. TRRS here is applied to the intercluster communication range in order to find the minimum communication range required for routing to take place. i.e. for network wide connectivity of clusterheads. Hence TRRS applied to HEED-ER follows the following steps: 1) Clustering formation as that performed in original HEED. 2) Sink broadcasts a HoTS-PL message to set the HoTS value of nodes in the network. 3) CMs send their reports to the corresponding CHs and CHs send reports < CHID >< data(optional) > to the sink at maximum intercluster transmission range. Data may be contained in the report message depending on the TRRS mode specified in the HoTS-PL message. 4) Sink obtains the total number of clusters in the network based on the number of reports it obtains. 5) Sink sends HoTS-PL reducing transmission range until the total number of clusters change.
978-1-4244-4148-8/09/$25.00 ©2009 This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.
6) Sink sends HoTS-PL increasing transmission range by one power level to restore original number of clusters. Fig. 6 shows the principle application of TRRS to HEEDER. We note that in contrast to DECRO-TRRS, TRRS application to HEED-ER does not require reclustering while the intercluster transmission range is being adjusted. This is due to the fact that clustering and routing are independent in HEEDER. Similarly, there are two options when performing the routing of data. For step 2, there is a choice of 1) reporting data to the sink simultaneously with finding the best range - TRRS with data or 2) first finding the best intercluster range, followed by data routing to the sink once the lowest transmission power that maintains complete network connectivity has been identified. These two approaches are termed TRRS(D) and TRRS in the simulations of Section IV. We note that the two modes of operation of TRRS have their respective advantage and disadvantage. The first option TRRS(D), eliminates the delay in finding the best transmission power as data are still being transmitted during the TRRS process. However the energy cost will be relatively high as CHs are transmitting data at higher transmission power than necessary. The second option of pure TRRS involves shorter messages from CHs only to discover the number of CHs reachable by the sink whilst the intercluster range is being decreased. CHs will route the messages by the same rules as when data was included in the packets. Although this method introduces the delay before transmission of data to the sink, it may more effectively save the energy of sensor nodes. Generally, if the amount of data to be sent is relatively large and sensor nodes have limited energy source, pure TRRS should be performed without data being sent during the TRRS process, and followed by reporting data at the final determined minimum transmission range. IV. S IMULATIONS The simulation is implemented in Java. An ideal MAC is assumed to exercise a MAC-independent protocol. Nodes initially have random data size and random energies. 4000 nodes are randomly distributed across 800 by 800 unit area. For DECRO and DECRO-TRRS, an initial range of 450 units is set. TRRS finds a best intercluster communication range of 380 units after 14 rounds of reclustering. At each round the cluster range is reduced by 5 units. DECRO proceeds with clustering at a clustering range of 450 units. Data aggregation factor is set to 50%. For HEED-ER, HEED-ER-TRRS and HEED-ER-TRRS(D), a clustering range of 390 units is used. The intercluster communication range is set to 800 units for all the three simulated protocols. A 50% data aggregation factor is used. HoTS messages are 16-bits, clustering Hello messages (e.g. data size, messages, CH claim message, GW claim message) are set to 32-bits, data sizes are between 0 - 8000 bits. HEEDER-TRRS(D) messages are 24 bits, and DECRO-TRRS(D) messages are 16-bits long. Energy consumption uses the model in [17], which uses 100 pJ/bit/unit2 amount of energy a node spends for broadcasts.
Fig. 7.
Fig. 8.
DECRO and DECRO-TRRS comparison.
HEED-ER and HEED-ER-TRRS comparison.
From Fig. 7, for the first 16 reports DECRO and DECROTRRS perform better than DECRO-TRRS(D). Pure DECRO does not expend energy in finding the best transmission range and this is reflected in the first initial reporting. Additionally the delay in finding the best range is omitted in pure DECRO. However the advantages are overshadowed after 16 reports and DECRO, performing at a higher transmission range than necessary uses up more network energy as shown in Fig. 7. DECRO-TRRS(D) which performs reclustering and data reporting whilst finding the best path has degraded performance for the first 16 reports, however since the best path is discovered, it outperforms pure DECRO after 16 reports. DECRO-TRRS(D), although it has the additional reclustering cost initially to find the minimum transmission range, once this transmission range is found, reporting can be done with the minimum required transmission range and power in oppose to pure DECRO which uses the higher transmission range throughout all the reporting. In DECRO-TRRS(D), the initial overhead of discovering the lowest required transmission range does not result in delay as the CHs are still broadcasting their aggregated data to the sink. The total energy includes energy consumed in HoTS initiation, clustering, and reporting (routing) of data to the sink. DECRO-TRRS which finds the lowest transmission range before data reporting outperforms
978-1-4244-4148-8/09/$25.00 ©2009 This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.
both pure DECRO and DECRO-TRRS(D) at the cost of the short delay in finding best transmission range (14 reclustering rounds without any data being sent). However if delay is not an issue in the startup of the system, DECRO-TRRS is the most effective approach in saving energy in the network as time progresses. In Fig. 8, HEED-ER-TRRS outperforms HEED-ER soon after the first report. As with HEED-ER the same intercluster transmission range is used throughout all consecutive reporting. HEED-ER-TRRS(D) does not use any additional overhead in discovering the lowest intercluster communication range, as this is done while data reporting is taking place during the discovery process. HEED-ER-TRRS first finds the minimum required intercluster range before initiating data reporting. Although HEED-ER-TRRS introduces the delay in initiating reporting, it outperforms HEED-ER and HEED-ER-TRRS(D) in regards to total network energy consumption. However it is important to note that the relative performance of TRRS and TRRS(D) is related to the number of rounds it takes to find the lowest required transmission range. Also no reclustering overhead is performed in both versions of HEED-ER-TRRS which shows its strong performance from initial reporting of data. We note that in regards to HEED-ER-TRRS, the messages broadcasted by the clusterheads to the sink are 16-bits, whereas the aggregated data messages sent by the clusterheads are much larger, as data sent to the clusterhead from each member of the cluster is up to 8000 bits and even after aggregation the relative sizes of packets are multifold. V. C ONCLUSION AND F UTURE W ORK In this paper we introduce the transmission range reduction scheme (TRRS) which aims at reducing power consumption in clustered wireless sensor networks. The application of TRRS was applied to two clustering algorithms, namely DECRO and HEED-ER, and its effectiveness was evaluated in regards to energy consumption. In both cases the effectiveness of TRRS was evident particularly when the rounds of data reporting to sink increases in time. We note that although the scheme reduces only the transmission power of nodes, even in the case that energy consumption is higher for reception, energy constraint devices still benefit by reducing the energy consumed in transmission in any communication network. This becomes more apparent in well or semi-synchronized networks, where nodes awake for data reception at the time a node is expected to transmit. Hence the multifold benefit of transmission power reduction can be applied to even low powered devices (
