1st International Workshop on GReen Optimized Wireless Networks (GROWN'13) Multichannel MAC for Energy Efficient Home Area Networks Kok Keong Chai, Shihab Jimaa, Yun Li, Yue Chen, and Siying Wang Abstract This paper proposes a multichannel medium access control (MAC) protocol for energy efficient IEEE802.15.4 home area networks that consists of novel allocation and superframe scheduling algorithm that improves the channel selection strategy and the superframe scheduling. It aims to improve the overall reliability and reduce the average delay. The proposed protocol is implemented in IEEE802.15.4 home area network andsimulation results show the proposed protocol outperforms the existing IEEE802.15.4 MAC protocols in the aspects of schedulability, reliability, delay and the overall throughput. Index Terms multichannel, MAC protocol, channel allocation, superframe scheduling I. INTRODUCTION uture home area networks (HANs) are expected to F contribute significantly towards improving energy usage and home comfort. In order for home users to determine the most cost efficient way to consume the energy within their homes, HANs provide the feedback of real time energy consumption and energy pricing.hans also enable all smart electronic devices to be connected together and enable automation control according to the environmental changes in the home and the preferences of the home user. However, HANs inherit most of the well-known challenges from traditional wireless networks, such as time varying channels and unreliable links. In additional to that, HANs also experience the challenges due to the diverse of devices and applications, higher application quality of service (QoS) requirements, limited transmission range, battery and computational power constraints, multi-hop operation, unidirectional links and synchronization issues. Kok Keong Chai, Yun Li, Yue Chen and Siying Wang are with the School of Electronic Engineering and Computer Science, Queen Mary University of London, U.K. (correspondence email: Michael.Chai@eecs.qmul.ac.uk) Shihab Jimaa is with the ECE Dept., College of Engineering, Khalifa University of Science, Technology and Research, UAE (email:saj@kustar.ac.ae). IEEE 802.15.4 is the standard for the low rate wireless personal area network (LRWPAN), which aims at the low speed, low energy consumption and low cost. It defines the specifications of physical (PHY) and medium access control (MAC) layers of the LR-WPAN [1].IEEE802.15.4 standardhas the capability of putting nodes into sleep mode periodically to save energy when there is no data to be transmitted. This characteristichas been widely adopted in many WPAN scenarios. Many literatures have researched into different MAC protocols and algorithms to adapt different requirements of wireless sensor networks. In [2], authors propose Superframe Duration Scheduling (SDS) algorithm that schedules different length of the superframe duration in the beacon-enable mode of IEEE 802.15.4 cluster-based networks. The algorithm sorts the superframes according to their beacon intervals and superframe duration. The major cycle is divided into several time slots whose length is the longest beacon interval and the superframe is put into the first available time slot. If all the superframes can find places to put, the set is considered as schedulable; otherwise, the scheduling fails and the set is not schedulable. Multichannel Superframe Scheduling (MSS) [3] is a multi-channel MAC protocol based on SDS. It modifies the SDS algorithm and adds the channel allocation method.in this scheduling technique, the personal area network (PAN) coordinator is required not only know the full network topology, but also the locations of allother coordinators. This means the network must be static and configuredbefore the networking start, or the PAN coordinator need to get the knowledge through some localization algorithms. However, locating all the coordinators will consume time and energy consumption while the results may not be accurate enough. Multi-frequency MAC (MMSN) protocol [4] designed specifically to wireless sensor networks. It specifies the frequency assignment and the media access method. In the MMSN, four frequency assignment methods are proposed: exclusive frequency assignment, the implicit consensus, even selection, and eavesdropping. 978-1-4799-0428-0/13/$31.00 2013 IEEE 28
In this paper, we propose a multichannel MAC protocol that applies novel multichannel allocation and superframe scheduling algorithm to improve the overall throughput and average frame delay. Simulation results show thealgorithm improves the superframe scheduling method as compared to SDS, and overcome the disadvantages of MSS and MMSN. The rest of the paper is organized as follows.section II explainstheproposed multichannel MACprotocol. In Section III, the simulation model and results are presented. Section IVgives the conclusions. II. PROPOSED MULTICHANNEL MAC The proposal multichannel MAC protocol in this paper is a distributed approach. Each coordinator makes their channel allocation decision locally using the information it receives. The proposed MAC protocol divides a main cycle into two windows, the channel reservation window and the superframe window. In the channel reservation window, only beacon can be transmitted. In this window each coordinator sends the superframe structure (the value of Superframe Duration (SD) and Beacon Interval (BI)), the channel decision of itself and its neighbors on the control channel within its time slot. And other coordinators choose their own channels and decide the superframe structure according to the information it receives. The superframe window is the data transmission period. In this window, coordinators communicate with their children using the reserved channel in the reservation window. A. Channel Reservation Window In this window, the time slots are separated with the equal number of coordinators. Every coordinator has a dedicated contention-free time slot to transmit its own beacon. Each coordinator derives the time slot it occupies using a unique ID. The first time slot is allocated to the node, which has the smallest ID, and similarly the last time slot is allocated to the node with the largest ID. Meanwhile, the parent should have the smaller ID than the children. So the PAN coordinator will occupy the first time slot. Each coordinator wakes up at the time slot of its parent and listens to the beacon of its parent and neighbors who transmit the beacons before it. The coordinator makes the channel selecting decision and decides the offset between the beginning of the transmission period and its own superframe according to the superframe structure and the channel decision of the nodes one and two-hops away. Then the coordinator sends its own beacon in its time slot including the superframe structure and the channel decision of itself and its parent and neighbors it has received. After its time slot, the node switches off the radio and goes into sleep and it won t wake up until the nextsuperframe window.the scheduling of the channel reservation window is illustratedin Fig. 1 The superframe structure of a coordinator is determined by its beacon interval and superframe duration. At the beginning of every cycle, a node can change its beacon interval and superframe duration according to the network condition and its own state, such as the number of children and the throughput, in the last cycle to adapt the network and improve the delivery efficiency. And this will also influence the channel selection of other coordinators. B. Superframe Window This window is divided into several minor cycles and every minor cycle is separated into two time slots of the same duration. In IEEE 802.15.4 standard, the beacon interval, BI, depends on the MACBeaconOrder (BO) and the SD depends on the MACSuperframeOrder(SO). (1) (2) Fig. 1 The channel reservation window 29
Fig. 2 Superframe structure Setting the time slot duration TS depends on the time slot order TO. (3) The SO of all coordinators must be equal to or smaller than TO and the BO of all the coordinators must be equal to or larger than the TO+1, so that the longest SD will not exceed a time slot and the smallest beacon interval is at least double the time slot, which is a minor cycle. So the duty cycle is 50% at most and lower duty cycle will save the energy. The PAN coordinator and the coordinators in even rank will put their superframe in the first time slot of each minor cycle and the odd rank counterpart will put their superframe in the second slot. In one channel, the superframe can be assigned next to each other only if it satisfies the sum of all the superframes duration is less than a time slot. In order to listen to the beacon of its parent and communicate with its parent, the children coordinators need to switch to the channel of their parent during the superframe of its parent. They will compete to access the channel with the nodes in the same cluster of their parent. The beacon interval of all the coordinators must be equal to or smaller than the major cycle of a superframe window. And after the superframe window, every coordinator adjusts its beacon interval and the superframe duration according to the network condition and its own state such as the number of children or the throughput). Then the channel reservation window in the next cycle starts.the superframe structure is shown in Fig2. C. Channel Selection Strategy The proposed algorithm uses all 16 channels that are available in IEEE802.15.4. The coordinator need to perform the channel selection for nodes within its cluster. Below is the proposed strategy: 1) The PAN coordinator occupies the first time slot in the channel reservation window. It chooses the channel first and by default, it chooses the first channel. 2) If a coordinator s superframe doesn t overlap with that of the others in one channel, then that channel will be selected. In order to decrease the switching cost between two time slots, the channel used by its parent has the priority to be chosen. 3) Foreach channel, the superframe can be assigned next to each other, only if it satisfies the sum of all the superframes duration is less than a time slot. 4) If a coordinator s superframe overlaps all others it can detect, another channel will be chosen. The channel is chosen randomly to improve the channel usage of all 16 channels. III. SIMULATION RESULTS A. Simulation Parameters To test the performance of the superframe scheduling algorithm, different amount of nodes are put in a home are network. The coordinators are selected randomly at the network initialization. The coordinators are divided into 4 levels while the first level only has a PAN coordinator. We assume all the nodes locatewithin a 100 m by 100 m squarehome, and the PAN coordinator is located in the center of the home. The packet size is 127 bytes, which is the maximum packet size defined byieee 802.15.4 standard. For each packet, the MAC frame occupies 121 bytes while the overhead added in physical layer is 6 bytes. The size of ACK frame is 11 bytes.the buffer size of a node is set to 51 frames, which means the buffer can holds maximum 50 frames. We considered an M/M/1/K packet model. The 16 channels in the 2450 MHz band are used for the transmission. B. Comparison with SMAC Sensor MAC (SMAC) [2] is one of the earliest MAC layer protocols for wireless sensor networks (WSNs). It 30
is the foundation of the MAC protocols for WSNs and it is widely used in many wireless sensor networks. We compare theproposed MACprotocol with SMAC in the aspects of reliability and the average frame delay. Fig.3 shows the reliability (successful packet sending rate) varies with the offered load of the system per node when there are 150 nodes in the home. From the simulation result, the reliabilityof SMAC drops sharply with increasing the offered load and when the offered load exceeds 0.5 frames per second the reliability of SMAC is less than 0.1. The reliability of the proposed MAC layer protocol isalways higher than that of SMAC and it decreases gradually as the offered load increases. C. Comparison with MMSN MMSN is a semi-dynamic and distributed multichannel MAC protocol for the IEEE 802.15.4. Each coordinator runs the algorithm locally toget the assigned channel number. However, its disadvantage is that, it needs a large number of channels when amount of nodes is large. The proposed MAC layer protocol is compared with the MMSN protocol in the aspects of reliability and delay. In this experiment, the algorithm is modified slightly byonly using 16 channels. So if the channel number thecoordinator computes is larger than 16, the coordinator will choose the least use channel. To showthat the performance of the proposed MAC layer protocol outperforms MMSN in the scenario of a largeamount of nodes, the simulations are conducted in scenarios of 150, 200, 250, 300, 350, 400 850 nodes. The offered load is fixedto 3 frames per second. Fig 5 shows the reliability of different number of nodes in the network. With the increasing number of stations, the reliability of MMSN drops rapidly. On the contrary, the reliability of the proposed MAC layer protocol only decreaseslightly. This is because as long as the number of nodes is in the schedulable range of the proposedmac protocol, the reliability will not decrease sharply since there is little interference between clusters due to collision. Fig. 3 The reliability under different load Fig.4 illustrates the performances of the average frame delay in the same scenario. Comparing with SMAC, the average wait time of the MAC layer is very small. Fig. 5 The reliability under different number of nodes And also, the proposed MAC protocol performs better than the MMSN in terms of average wait time as shown in Fig.6. The average waiting time of the proposed protocol has a relevant stable performance with the increasing number of stations. Fig. 4 The average wait time under different load 31
The simulation results indicated that the proposed multichannelmac layer protocol is superior to the SDS, SMAC and MMSN protocols. Meanwhile, the proposed multichannel MAC protocol can solve both the direct and indirect beacon collision problems. Moreover, the channel allocation and superframe scheduling algorithm is a distributed and semi-dynamic method which means the protocol is suitable for thedynamic network. Unlike some existing multichannel protocols, the proposed multichannel MACprotocol doesn t require special hardware such as multiple radios. So the protocol has a good physical reliability. Fig. 6 The average wait time under differentnumber of nodes IV. CONCLUSIONS In this paper, a new multichannel MAC layer protocol for IEEE802.15.4 networks is proposed. The protocol divides a main cycle into two windows. Every coordinator has a contention-free time slot to send the scheduling information of itself and its parent and neighbors. Meanwhile, it wakes up at the time slot of its parent and receives the scheduling information of its parent and neighbors, and then it can make the decision of channel allocation and superframe scheduling itself locally using the information. If the coordinator cannot find a free time slot to put its superframe without interference, it will choose another channel. After its own time slot, the coordinator will sleep until the superframe window. In the superframe window, the coordinators communicate with their children using the channel reserved in the reservation window. As for the transmission within thecluster, the CSMA/CA mechanism is adopted. REFERENCES [1] IEEE std. 802.15.4, Part. 15.4: Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Low-Rate Wireless Personal Area Networks (LR-WPANs), IEEE Std., 2011. [2] Koubaa, A., Cunha, A., &Alves, M. A time division beacon scheduling mechanism for IEEE 802.15. 4/Zigbee cluster-tree wireless sensor networks. In,ECRTS'07. 19th Euromicro Conference onreal-time Systems, 2007, pp. 125-135 [3] Toscano, E., & Bello, L. L. Multichannel Superframe Scheduling for IEEE 802.15. 4 Industrial Wireless Sensor Networks. Industrial Informatics, IEEE Transactions on, 2012, 8(2), pp.337-350. [4] Zhou, G., Huang, C., Yan, T., He, T., Stankovic, J. A., &Abdelzaher, T. F. MMSN: Multi-frequency media access control for wireless sensor networks. In INFOCOM 2006. Twenty-First Annual Joint Conference of the IEEE Computer and Communications Societies. Proceedings, April,2006,pp.1-13. [5] Ye, W., Heidemann, J., &Estrin, D. An energy-efficient MAC protocol for wireless sensor networks. In INFOCOM 2002. Twenty-First Annual Joint Conference of the IEEE Computer and Communications Societies. Proceedings. Vol. 3, pp. 1567-1576. 32