An Ultra-low-power Medium Access Control Protocol for Body Sensor Network

An Ultra-low-power Medium Access Control Protocol for Body Sensor Network Huaming Li and Jindong Tan Department of Electrical and Computer Engineering Michigan Technological University Houghton, MI 49931, USA Abstract In this paper, BSN-MAC, a medium access control (MAC) protocol designed for Body Sensor Networks (BSNs) is proposed. Due to the traffic coupling and sensor diversity characteristics of BSNs, common MAC protocols can not satisfy the unique requirements of the biomedical sensors in BSNs. BSN-MAC exploits the feedback information from the deployed sensors to form a closed-loop control of the MAC parameters. A control algorithm is proposed to enable the BSN coordinator to adjust parameters of the IEEE 82.15.4 superframe to achieve both energy efficiency and low latency on energy critical nodes. We evaluate the performance of BSN-MAC by comparing it with the IEEE 82.15.4 MAC protocol using energy efficiency as the primary metric. duty cycle schedule between sensor nodes. S-MAC introduces virtual clusters to enable nodes to communicate within divided time slots according to the exchanged schedule. S-MAC trades off energy for latency. S-MAC, T-MAC and D-MAC all suffer from synchronization overhead and periodic exchange of sleeping schedules. Moreover, all of them are typically designed for multi-hop ad hoc wireless sensor network, while in this paper our main concern is a star-topology BSN. Keywords- body sensor network (BSN); medium access control (MAC); IEEE 82.15.4. I. INTRODUCTION The last decade has witnessed a significant progress in wireless sensing/monitoring and wearable/implantable biosensors for healthcare, which arises the interest in building a network to connect the sensors wirelessly. Using wireless communication instead of traditional wired connection can reduce the maintenance cost of healthcare systems, make the ubiquitous and mobile healthcare possible, and give the patients more freedom and comfort. However due to the implanted feature, replacing these sensors or charging their battery is extremely difficult and often needs a surgery operation. Body Sensor Networks (BSNs) are networks whose nodes are either implanted or close to human bodies. BSNs specially address the needs of low-power wireless communication between wearable or implantable biosensors. As shown in Fig.1, a BSN usually consists of implantable or wearable biosensors, such as glucose sensors, blood pressure and oxygen saturation (SpO 2 ) sensors, temperature sensors and even ingestible camera pills. These sensors continuously monitor vital signs and report data to a powerful external device, such as a PDA, cell phone or bedside monitor station. Medium Access Control (MAC) plays an important role in determining the energy consumption in wireless communication, and hence, the lifetime of a node. In the traditional contention-based MAC protocols, low power listening [1] and preamble sampling [2] are used to reduce the idle listening time by periodically turn off the radio to reduce duty cycle. S-MAC [3], T-MAC [4] and D-MAC [5] are proposed to solve the idle listening problem by applying a synchronized Figure 1. A body sensor network consisting of various biosensors ER-MAC [6] introduces the concept of energy-criticality of sensor nodes. It is a function of energies and traffic rates. Depending on the energy-criticality, this protocol makes more critical nodes sleep longer hence to balance the energy consumption of the whole network. In [7], the authors present B-MAC employing an adaptive preamble sampling scheme to reduce duty cycle and minimize idle listening. Its core part is to use noise floor estimation to achieve accurate channel clear assessment (CCA). B-MAC solves the false positives from which other protocols are suffering as a source of the idle listening. B-MAC is designed to be flexible and provides several interfaces to enable other services to change the behavior of B-MAC. Therefore, it leaves the adaptive control work to other services. In this paper we propose BSN-MAC (Body sensor network MAC), which is a dedicated ultra-low-power MAC protocol designed for star topology BSNs. BSN-MAC has good compatibility with IEEE 82.15.4 as well as accommodates unique requirements of the biosensors in BSNs. BSN-MAC is designed to be an adaptive MAC protocol. By exploiting feedback information from distributed sensors in the network, BSN-MAC adjusts protocol parameters dynamically to achieve best energy conservation on energy critical sensors. -783-9152-7/5/$2. (c) 25 IEEE

The rest of the paper is structured as follows. First, we introduce the background of BSN-MAC. Then, we present the information-aware BSN-MAC in Section 3, including subsections about the discussion of medium access priority and the adaptive algorithm in detail. In Section 4 we give the simulation result. Finally, we present the conclusions. II. BSN-MAC BACKGROUND BSN-MAC is based on IEEE 82.15.4 which supports both star and peer-to-peer network topologies. Here we concentrate on the star topology because in a BSN the number of sensors is limited and an external mobile device such as PDA, cell phone is usually available. This external device may act as a BSN coordinator and a gateway to other networks. Since it is relatively easy to charge or replace the battery of external devices, they can be considered as not restricted by power and computing resources. Therefore, it is possible to make a tradeoff between the energy of the coordinator and the lifetime of the energy-critical implanted sensor nodes using an adaptive MAC protocol. Figure 2. IEEE 82.15.4 superframe structure The MAC protocol in IEEE 82.15.4 can operate in both beacon and non-beacon modes. We will put our focus on beacon-enabled mode, in which the BSN coordinator controls the communication by sending out regular beacons in a superframe for network synchronization and network association control. The BSN coordinator defines the start and end of a superframe by transmitting a periodic beacon, as shown in Fig. 2. The overall superframe length (BI) and the active superframe duration (SD) are specified by two parameters respectively: macbeaconorder (BO) and the macsuperframeorder (SO) [8]. BI = abasesuperframeduration 2 BO (1) SD = abasesuperframeduration 2 SO (2) BO and SO can be chosen from to 14 for beacon mode. Since active superframe duration (SD) is included in overall superframe length (BI), we have SO BO 14 (3) When BO = 14 and SO =, the lowest duty cycle 1/16384 can be achieved, with the beacon interval at around 252ms. By default, the active portion of a superframe is evenly divided into 16 time slots and contains three parts: the beacon, a contention access period (CAP) and a contention free period (CFP). III. INFORMATION-AWARE ADAPTIVE BSN-MAC A. Medium Access Priority Depending on the different applications, we categorize the sensors in BSNs into four types based on their energy critical level: unconstrained, constrained, highly constrained and extremely constrained. Because the battery of a mobile gateway (BSN coordinator), like a PDA or cell phone, is rechargeable or replaceable, its energy can be considered as unconstrained. The energy of wearable sensors is classified as energy constrained. Ingestible sensors will enter human bodies, therefore are considered as highly constrained. The energy for implanted sensors is extremely constrained. In the BSN-MAC protocol, three types of devices are assigned with different priorities to access the medium in order to guarantee the ultralow-power consumption on extremely energy constrained implanted devices. Besides the energy constraint level, we also consider the following factors to decide the medium access priority: sensor remaining energy level, sensor remaining buffer level and sensory data time criticality. In BSN-MAC, we formulate the sensor medium access priority as: P MAC = 1 2 EL + TC E R B Here EL is the energy constraint level., 1, 2 and 3 are assigned to energy unconstrained, constrained, high constrained and extremely constrained sensors respectively. T C is the criticality index of the sensory data. To send out data of an abnormal heart rate in time may be vital to a patient, while to transmit common body temperature data is relatively trivial. T C is chosen from 1 to 3 representing low, normal and high time criticality. E R represents the energy remaining level and is chosen from 1 to 3. A larger number means higher remaining energy level and may decrease the medium access priority of a sensor. B R represents the remaining buffer level of a sensor and is assigned from 1 to 3. A larger B R means larger remaining data buffer and also decreases the medium access priority since with the buffer a sensor can endure more latency without losing any data. P MAC is used in BSN-MAC to provide a guideline to assign different medium access priorities to different sensors. B. BSN-MAC Design To achieve ultra-low-power performance, we make two tradeoffs in BSN-MAC design. The first is between the energy of coordinator (unconstrained) and the life of energy-critical sensors nodes. The second is between the sensor data report delay and the life time of energy-critical sensors nodes. First, we introduce a hybrid synchronization scheme (Fig. 3). In this hybrid scheme, the sensors power off their radio most of the time and only wake them up when there are interesting data need to be transmitted. Once the nodes wake up and get synchronized, we let the sensors keep synchronized by continuously tracking the beacon frames until the entire data transmission is accomplished. After the transmission, R (4)

sensors power off their radios again. Thereby, a sensor only needs to pay for the first round idle listening cost when transmitting several packets. Through this hybrid synchronization scheme, we avoid duplicate idle listening time when there are several packets to be transmitted. In the meantime, we also avoid receiving the unnecessary beacon frames. Sleep Sensor Node Radio on FEEDBACK: Energy constraint, time criticality, buffer level BSN Coordinator Adaptive Parameters: BO, SO, GTS, BLE Figure 3. The hybrid synchronization and adaptive superframe Sleep mode Inspired by the feedback mechanism widely used in control theories, BSN-MAC exploits the feedback application-specific and sensor-specific information from the nodes to achieve dynamic MAC parameter adaptation. After the network association we let the chosen coordinator (usually a pocket mobile device) broadcast beacon frame using BO = SO =. This puts the duty cycle of the coordinator to 1% and gives the shortest beacon interval. When sensors wake up and try to transmit data, this shortest beacon interval reduces synchronization idle listening time to the minimum. Here we sacrifice the coordinator to reduce power consumption of the sensor nodes. We always keep SO and BO to be the same to provide the largest bandwidth and lowest latency. Once the node powers up its radio and gets synchronized, depending on the beacon information and buffered data size, it either sends out a data transmission request (REQ) packet followed by part of the sensory data or sends out the sensory data directly. After the node synchronizes with the coordinator, by decoding the beacon frame, the node learns the length of the beacon interval and understands whether it can transmit its data during a single beacon interval. If yes, it will start transmitting its data directly. Otherwise, the node sends out a REQ packet first to ask the coordinator to change the beacon interval. The REQ packet includes the P MAC (medium access priority, as shown in (4)) of the sensor, which is used by the network coordinator to determine the priority when multiple REQs are received. The sensor can not always change the MAC parameters as their will since when the MAC parameters are changed to favor a certain node, other nodes may suffer from it. P MAC is determined by the energy constraint level, data time criticality and the remaining energy and buffer level as shown in (4). We notice that all the information needed to calculate P MAC is available on the sensors side. We put all the computation on the sensors side and let the sensors send back the calculated P MAC value only instead of all the sensorspecific information. This method effectively reduces the control overhead at the cost of some affordable computation. In addition to P MAC, we also include RBI (Recommended Beacon Interval) and RBN (Recommended Beacon Number) in the REQ packets. RBI and RBN are calculated based on P MAC, the network condition and the size of data to be transmitted. Following the same design idea, the computation is put on the sensor s side also. RBI is calculated as 2 RBI S S S Base Data Base = (5) S Base is the amount of data that can be transmitted during a single beacon interval when BO=SO=. S Data is the data size to be sent. We subtract S Base from S Data because the sensors transmit the sensory data after the REQ packet before the coordinator takes any change in the beacon interval. RBI is rounded up to an integer. Using RBI sensor data are supposed to be sent within a single beacon interval which reduces the overhead and radio power up time to the minimal. If RBI is used, every node becomes extremely selfish and tends to conserve its energy as much as possible. However, as we mentioned earlier, sensors can not always change the MAC parameters at their will for fear of affecting other nodes badly. BSN-MAC requires sensors to take P MAC into consideration when deciding the final RBI. From (4), we can see the range of P MAC is from 2/9 to 48. We use (6) to determine the final RBI. =,( > 24) = /2,(24 > 8) = /4,(8 > 2) = /8,(2 2/9) Therefore, only the nodes with the highest medium access priority can request the coordinator to change the beacon interval freely. All the other sensors may change the beacon intervals, but not that significantly. RBN is used by the sensors to describe how many such requested beacon intervals they need to transmit the current data. We point out that part of the sensory data is also sent following the REQ packet within the same beacon interval. In this way bandwidth waste is reduced to minimum. The sensor side algorithm is described in Algorithm 1. S THRSHOLD is the size of data that can be sent during the current beacon interval. Algorithm 1 MAC parameters adaptation (Sensors) 1: While no stop command do 2: If no data to transmit then 3: Power off radio 4: Else 5: Wake up radio and receive beacons 6: Set S THRSHOLD according to the beacon frame 7: Set CBI current beacon interval 8: If S DATA S THRSHOLD then 9: Transmit data directly 1: Else 11: Calculate P MAX according to Eq. (4) 12: Calculate RBI 13: If RBI > CBI Then 14: Calculate RBN 15: Send P MAX, RBI and RBN in REQ (6)

16: Send Data 17: Else 18: Transmit data directly 19: End If 2: End if 21: End While The network coordinator sends out beacons with BO=SO= at the beginning. Once the coordinator receives a REQ packet, it compares the P MAC value in the REQ packet with the current recorded value CP MAC, only if P MAC is larger than CP MAC the coordinator changes the network parameters according to the requested value. The coordinator also maintains a count CRBN for the requested beacon number. If the CRBN is reduced to or the coordinator detects an idle channel duration that exceeds a threshold T I, the coordinator changes the BO and SO to to guarantee the sensors can get synchronized rapidly. The coordinator algorithm is shown in Algorithm 2. Algorithm 2 MAC parameters adaptation (Coordinator) 1: Set CRBN 2: While no stop command do 3: If CRBN = or Idle time > T I then 4: Set BO, SO, CP MAX 5: End if 6: Receive packets 7: If REQ packet then 8: Read P MAC, RBI and RBN 9: If P MAC > CP MAC then 1: Set BO RBI, SO RBI, CRBN RBN 11: Else 12: Keep current SO and BO 13: End if 14: End if 15: Send beacon frame with current BO and SO 16: If CRBN > Then 17: CRBN CRBN 1 18: End if 19: End While IV. PERFORMANCE EVALUATION We implement our algorithms in NS-2 network simulator (2.27) with the IEEE 82.15.4 WPAN simulation extension from Zheng [9] of Samsung/CUNY. This extension package provides most of the IEEE 82.15.4 MAC and PHY layer functionalities required by BSN-MAC. The simulation is performed in a beacon enabled star environment. 7 nodes are selected to form the body sensor network with the coordinator (red circle) sitting at the center. The radius of the star network is chosen as 1m, and the transmission range is 15m. Although for a body sensor network, most of the sensors may locate on the human body and work within a limited area, we choose this relatively big transmission range and network size with the consideration of potential increased requirements. In the simulation, we use different constant bit rate traffics to simulation the behaviors of different sensors. First, we let the sensors transmit a certain amount of data continuously using different packet sizes and different beacon intervals to see the importance of the adaptive MAC parameter in BSN- MAC. We use 1MB data in the simulation and assume the sensors have got synchronized with the coordinator successfully. BO increases from to 6 and BO and SO are always chosen to be the same to provide the maximum bandwidth. To eliminate the affection of routing and network association, we begin to calculate the energy consumption when the network is stable and the routing path has been established. 45 4 35 3 25 2 15 Packet size 1 Packet size 5 Packet size 25 1 1 2 3 4 5 6 Figure 4. Energy cost for transmitting 1MB data with different beacon orders and packet sizes 6 5 4 3 2 1 45 4 35 3 25 2 15 1 5 Buffer size 512B Buffer size 1KB Buffer size 4KB Buffer size 8KB 1 2 3 Buffer size 512B Buffer size 1KB Buffer size 4KB Buffer size 8KB (a) 1 2 3 4 5 6 Figure 5. Energy cost for transmitting 1 MB data with different buffer (packet size = 1B). (a) BO from to 3, (b) BO from to 6 The simulation result in Fig. 5 shows shat using larger beacon interval and larger packet size can reduce the energy cost effectively. The benefit mainly comes from the reduced radio power up energy consumption and transmission (b)

overheads which include packet overheads and extra beacon frames. In addition, we compare the energy cost when transmitting 1MB data with different transmission patterns. This experiment aims to evaluate the performance of different sensors operating under BSN-MAC. Different buffer sizes of the sensors bring different transmission requirements. We assume the sensor buffer is full before each transmission and 1B packet size is used to reduce the packet segmentation to the minimum. Four different buffer sizes: 512B, 1kB, 4kB and 8kB are used in the simulation. The hybrid synchronization is used. Sensors power up their radio and get synchronized to transmit data when their buffer is full. Once the buffer data are transmitted, the sensor powers off their radio to conserve power. From Fig. 5(a), we can see that for transmitting the same amount of data the sensors with a smaller buffer consume more energy comparing with the sensors with a larger buffer. The main performance difference comes from the different idle listening time for the synchronization. Every time the sensor powers up radio to transmit data, it may spend BI/2 idle listening time in synchronization with the coordinator. Due to the different buffer size, the sensors with larger buffer can power up their radio less times for transmitting the same amount of data comparing with the sensors with smaller buffers. Thus, the sensors use larger data chunk may reduce the synchronization cost. Figure 6. 5 45 4 35 3 25 2 15 1 5 BSN-MAC BO fixed at BO fixed at 3 2 4 6 8 1 12 14 16 18 2 Transmitted data size [MB] Energy cost comparation between BSN-MAC and static IEEE 82.15.4 MAC protocol With the beacon order increasing, the idle listening penalty increases exponentially and becomes the dominant factor of energy cost, shown in Fig. 5(b). The rapidly increasing idle listening cost counteracts the benefit from the reduced overheads and makes the overall energy increase dramatically. These results justify our algorithm in BSN-MAC. From Fig. 4 we can see using a large beacon interval to transmit data continuously can improve energy efficiency. However, a large beacon interval makes the sensors suffer from a high idle listening penalty. This contradiction can not be solved with a static MAC protocol. In BSN-MAC we use adaptive beacon intervals to avoid the heavy idle listening penalty while let the sensors enjoy the benefit from the reduced overhead safely. In Fig. 6, we compare the performance of dynamic BSN- MAC and the static IEEE 82.15.4 MAC. In this simulation, we choose a buffer size of 4KB. In the BSN-MAC, network beacon intervals are changed dynamically to achieve best energy efficiency. In the static IEEE 82.15.4, we fix the beacon interval at and 3 and assume the hybrid synchronization scheme is used to save power. We choose 3 as one of the static beacon intervals because this is the optimal beacon interval BSN-MAC use in the transmission. In this simulation BSN-MAC outperforms the static IEEE 82.15.4 with BO= by 23% and outperforms the static IEEE 82.15.3 with BO=3 by over 4%. We should notice if we do not include the hybrid synchronization scheme in IEEE 82.15.4, the difference would be more significant. V. CONCLUSTIONS In BSN-MAC, we adopt a cross layer design strategy considering the information obtained from the sensors to guide the medium access. The network coordinator and the sensors interact to achieve efficient power management. A hybrid synchronization scheme is used. The coordinator controls the communication by varying the superframe structure. The sensors provide real-time feedback to BSN coordinator with application-specific and sensor-specific information, such as sensor types, remaining data buffer and energy levels and data transmission rate requirements. Hence, the BSN coordinator can make dynamic adjustment based on the feedback to achieve better performance in energy efficiency and latency. Simulation results show the MSN-MAC can improve the energy efficiency and prolong the sensors life time. REFERENCES [1] J. Hill and D. Culler. Mica: a wireless platform for deeply embedded networks. IEEE Micro, 22(6):12 24, November 22. [2] A. El-Hoiydi. Aloha with preamble sampling for sporadic traffic in ad hoc wireless sensor networks. In Proceedings of the IEEE International Conference on Communications (ICC), New York, April 22. [3] Wei Ye, John Heidemann, and Deborah Estrin. An Energy-Efficient MAC Protocol for Wireless Sensor Networks. In Proceedings of the IEEE Infocom, pages 1567 1576, New York, NY, USA, June 22. [4] T. van Dam and K. Langendoen. An adaptive energye-fficient MAC protocol for wireless sensor networks. In Proceedings of the 1 st ACM Conf. on Embedded Networked Sensor Systems (Sen-Sys), pages 171 18, Los Angeles, CA, November 23. [5] Gang Lu, Bhaskar Krishnamachari, and Cauligi Raghavendra. An Adaptive Energy-Efficient and Low-Latency MAC dor Data Gathering in Sensor Networks. In Proceedings of the 4 th International Workshop on Algorithms for Wireless, Mobile, Ad Hoc and Sensor Networks, April 24. [6] Rajgopal Kannan, Ram Kalidindi, S. S. Iyengar, Vijay Kumar. Energy and rate based MAC protocol for wireless sensor networks. Special section on sensor network technology and sensor data management, Volume 32, Issue 4, Pages: 6 65, December 23. [7] Joseph Polastre, Jason Hill, David Culler. Versatile low power media access for wireless sensor networks. In proceedings of the 2nd international conference on Embedded networked sensor systems. Pages: 95 17, 24. [8] IEEE 82.15.4, Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Low-Rate Wireless Personal Area Networks (LR-WPANs), IEEE, October 1 23. [9] Jianliang Zheng and Myung J. Lee. Will IEEE 82.15.4 Make Ubiquitous Networking a Reality?: A Discussion on a Potential Low Power, Low Bit Rate Standard. IEEE Communications Magazine, 42(6):14 146, June 24.