Tbcd-tdm: novel ultra-low energy protocol for implantable wireless body sensor networks

TBCD-TDM: Novel Ultra-Low Energy Protocol for Implantable Wireless Body Sensor Networks F. Fereydouni Forouzandeh1, O. Ait Mohamed1, M. Sawan2, F. Awwad3 1 Electrical & Computer Eng. Dept., Concordia University Montreal Canada 2 Electrical Engineering Dept., École Polytechnique Montréal Canada 3 College of Information Technology, United Arab Emirates University {f_fereyd,ait}@ece.concordia.ca Abstract— The field of Remote health monitoring now includes technologies such as home and mobile health monitoring, teleretinal imaging, tele-radiology, remote cardiac monitoring, video conferencing and sensors for remote diagnosis and treatment to patients. In this regard, implantable wireless body sensor networks (IWBSNs) have recently emerged as an important and growing research area. These implantable sensors are required to be reliable, very small, battery-operated, and capable of collecting data, processing it, and transmitting it wirelessly and efficiently. Since these devices are required to run with limited resources (energy, processing, and memory), their utility protocols (collecting, processing, and communication) should be designed carefully, not only to work reliably but, more importantly, to be resource-efficient. The life time of the embedded batteries associated with these sensor nodes varies from a few days to a few weeks as was described in a previous work by the authors. In this paper, we propose a novel technique which allows the implanted sensor nodes to communicate with a base station located outside the body efficiently by consuming the minimum amount of energy. Our proposed protocol allows the battery to last significantly longer even for years with a gain of up to 100’s times of power saving. This will improve the quality of patient life, and reduce risk of infection resulting from frequent chirurgical operations needed to replace such implantable batteries. Also, a new time synchronization algorithm is briefly introduced in this work that is especially applicable to our proposed communication protocol. Index Terms— Bioelectronics, Implantable biomedical devices, Low-energy, Protocols, Wireless Body Sensor Networks.

T

I. INTRODUCTION

HE advances in medical chirurgical techniques allows the implantation of tiny sensor nodes inside the human body in order to gather appropriate information from different spots or parts of the body to monitor its physiological changes for health care as well as for research studies purposes. These implantable sensor nodes should be designed as small as possible so that the body does not reject them and hence to stay well-functioning inside the body for several years. They are also wirelessly connected to a base station forming the system called as; Wireless Body Sensor Network (WBSN), Fig. 1. Information such as heart rate and status (ECG),

Implanted Body Sensors Data

Data

Node-1 Node-2

Short Signals

External to Body Base Station

Data

Node -i

Data

Node-n

Distribute all IDs and Data

Fig. 1. WBSN model, (Several Wireless Sensors + Base Station).

brain status, body temperature, blood urine, blood glucose and blood pressure are to be gathered, processed and wirelessly transmitted to a local Base Station (BS) outside the body. Each sensor node consists of four major components, a battery, a sensor, a controller and a transceiver including a tiny antenna. The transmission type and the transmission protocol play a major role in the longevity of the sensor battery. Since the radio transmission consumes huge amount of energy, the definition of new techniques and protocols, with savage energy capability, are critical for the reliability of these WBSNs. In [1], we defined an ultra-low-energy protocol at a higher level of abstraction and we showed its correctness theoretically. In this paper, we present several enhancements of this protocol, named as: Time-Based Coded Data-Time Division Multiplexing (TBCD-TDM), demonstrating its effectiveness and correctness through simulations. The protocol would allow then saving the very small amount of energy available in the very tiny battery of the sensor node hence lasting for longer period of time inside the human body. Our results are compared as well to other known protocols in this area of research such as ZigBee [2]. In addition, we show that our proposed protocol should be implemented using the Amplitude Shift Keying (ASK) technique instead of the Frequency Shift Keying (FSK). The rest of the paper is structured as follows. Section II surveys the current relevant standard protocols along with the power consumed by the transceivers. In Section III, new low-

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.

energy-consumption communication protocol for wireless body sensor nodes is presented. The simulation results are demonstrated in section IV. An analysis to the time synchronization is given in section V. Comparison and the suitable modulators for such applications are discussed in section VI, and finally section VII concludes the paper with future work. II. RELATED COMMUNICATIONS PROTOCOLS TinyOS, is an old protocol which is good for wearable BSNs, but not for the implantable [3]. Tree-Based (TB) and Cluster-Based (CB) protocols in [4] use only a small number of nodes to transmit data to the external base station. LEACH (Lower Energy Adaptive Clustering Hierarchy) protocol in which its main techniques are based on algorithms for distributing cluster forming, adaptive cluster forming, and cluster header position changing, is described in [5] and [6]. These algorithms ensure fair energy dissipation among all sensor nodes prolong the lifetime of the sensor nodes. In LEACH, time is partitioned into several intervals with equal length and runs with many rounds. Each round contains two states: cluster setup state and steady state. In cluster setup state, it forms cluster in self-adaptive mode; in steady state, it transfers data. A sensor dies once its battery dies. The simulation results have shown that the LEACH can extend the sensor network lifetime up to eight times longer [6]. Another proposed protocol in this area is PEGASIS (Power-Efficient Gathering in Sensor Information Systems) [7], which is an improved version of LEACH. In PEGASIS, each node communicates only with a close sensor node and takes turns transmitting to the base station. Based on the simulation results for this method, PEGASIS performs about 100% to 300% better than LEACH when 1%, 20%, 50%, or 100% of nodes die [7]. All of these protocols use the same standard methods which usually play with the topology of the network, and utilize different compression methods. They attach huge amount of control bytes to the main data to be sent on each round of communication. For example, ZigBee, which is one of the best and commonly used protocols in sensor networks, adds about 27 control bytes to the main data (12 bytes for request and 15 for confirmation) [2]. This keeps the transceivers active for a long period of time. In the next section, we will show how our new proposed protocol leads to save energy from the power source. This is achieved by keeping the transceivers in sleep mode for long periods of time and in active mode for very short periods of time with a very small duty cycle which is about 0.001% to 0.005% versus at least 0.1% in the case of the ZigBee standard.

minimum activity status, which is the sleep mode almost all the time. Let us start by considering three sensor nodes having equal data ranges. We assume that all sensor nodes are perfectly synchronized with the BS. This is achieved by using the same clock frequency for the BS counter and the counters within each sensor node. In addition, all counters should be initialized with the same value. The transceiver in the BS is supposed to be always in active mode, but the sensor nodes will always be kept in sleep mode except when the value read by the sensor coincides with the counter value. At this particular event, the sensor wakes up to send a very short signal and then quickly return back into sleep mode. This will significantly help the sensor node to prevent wasting large portion of energy available in its tiny battery. These events are generated by the ID_Counter for identifying the sensor, and the Data_Counter to find the corresponding data for that sensor. In fact, a single counter is used in this design that is splitted in two portions: the lower order is called Data_Counter and the higher order is ID_Counter. The BS and all sensor nodes are always reading the same value from these counters using a synchronized clock input. This recalls the property of Time Division Multiplexing (TDM), however a full data or packet will never be sent. The structure of these sensor nodes and the BS are shown in Fig. 2 followed by the description of the algorithms at the sensor node and the base station node. The frequency of this clock is based on the resolution of the data range to be sampled by the sensors. For instance, if there is a range of 8 different levels to be sampled by the sensor, then the width of the data counter would be 3-bits. Obviously, whenever the unsigned value in the data counter overflows, an ID_Counter event occurs by incrementing by one for all nodes. Then, Data_Counter restarts counting from zero to catch the sensor data for the current node (ID). The zero status of the ID_Counters is reserved for initialization and configuration phase.

Sensor Data

(a)

DTx To Workspace5 Random Integer

Counters

Random Integer Generator 1

Clk

Data

=

==

Hit

Rst

Verify _Data

Data _Counter 1 ID

Bitwise AND

Clock

Synch.

Clk

ID

Up

Rst

Cnt

= ==

ID

Step 2

ID_Counter

Verify _ID

Clk

PROPOSED PROTOCOL

Pulse Generator 1

In this work, instead of transmitting the value of the sampled data in each round, the data is coded in a time domain range and only a very short signal will be sent to the BS by the transmitter sharply at a corresponding level within this range. This will significantly extend the battery lifetime of the sensor node in human body by keeping the transceivers in their

1 Tx

Generate Tx

RxBS

Counters

III.

Send pulse

Cnt

Up

Cnt Up Data

Rst

To Workspace2 In1

Out1

(b) 1 Sensor_Data

Data

Sample _Data

Hit

Data _Counter 1

ID

Rx

Receive Pulse

Synch.

Clk Rst

Reset

Up Cnt ID

2 Sensor_ID ID _BS

ID_Counter

To Workspace 3

Fig. 2. a- Sensor model, b- Base Station model.

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.

General Algorithm for Sensor Nodes: 1234567891011-

Initialization and synchronization, Assign Node_ID Reset counters, Start counting ID & Data counters, (Clk : f KHz) Start: if ID_Counter = 0 then do configuration and updates else wait until ID_Counter = Node_ID wait until Data_Counter = Sensor_Data (through ADC) Turn On the Transmitter Send Short Signal Turn Off the Transmitter end if loop Start // repeat for the next round

Node Sensor-1

Sensor-2

Sensor-3

General Algorithm for Base Station: 123456789-

BS

Data Counter

Initialization and synchronization, Assign all Node_IDs Reset Counters, Start counting ID & Data Counters (Clk: f KHz) Start: if ID_Counter = 0 then do configuration and updates else Wait on Short Signal from the sensor Obtain Sensor_ID from ID_Counter Obtain Sensor_Data from Data_Counter end if loop Start // repeat for next round

Therefore, for instance a 4-bit ID_Counter covers a maximum of 15 sensors within the range. If a value in Data_Counter of node (i) hits its sensor data through an ADC (Analog to Digital Converter), it immediately turns its transmitter ON and generates a short signal on the antenna. At the receiver side of the BS, this data would be easily obtained through a DAC (Digital to Analog Converter) corresponding to that data of sensor node (i). Assuming that sensor “i” sends the data “d” to the base station, the key properties of the protocol are given below:

ID Counter 0

0.5

1.5

2

2.5

Zoom

Time: Sec

3

39 Æ

40 Æ

38 Æ Data 42 41 40 39 38 37 36 Offset , 35

7 6 5 4 3 2 1 Æ 0

Æ Æ Æ

ID-3

Sensor node side:

For each sensor i

ID-2

R (DATA_S , ID_S ) AND exists d. (DATA_S =d) AND (ID_S = i)

ID-1 Initial 0.92

Implies data_valid(i) = 1.

0.93

0.94

0.95

0.96

0.97

0.98

0.99

1

Sec

Fig. 3. Simulation example of three Sensors with 3-bit data.

Base Station side: Exists i. Data_valid(i)=1 AND ID_bs = i implies DATA_bs = d

where R is the relationship between all counters in sensor nodes and the BS and synchronization, DATA_s and ID_s are counters in sensor nodes while DATA_bs and ID_bs are counters in the base station. IV. VALIDATION The proposed protocol has been modeled and simulated in Windows version of MATLAB_SIMULINK, and the functionality of this protocol is verified from very low to very high resolution steps. For instance, the simulations results in Fig. 3 shows the results obtained from only three temperature sensor nodes with the resolution of 8 levels per each to read the body temperature in range 35 to 42 degrees. There is one

sine-wave generator per sensor used as input source data, each with different frequency. The obtained results in the BS side for each sensor node are shown by the stepped sine-wave over each corresponding sensor source. The ID_Counter and Data_Counter are also shown assuming all synchronized in all sensor nodes with the BS. The magnified portion of the Fig. 3 is clarifying the results in a visible resolution. The 2-bit ID_Counter determines to which sensor node the obtained data belongs. The 3-bit data counter (with 8-step resolution and an offset value of 35) produces the corresponding data in the range of 35 to 42 once the short signal is detected on the RX output of the receiver in the base station. For example, the sensor node #2 (with ID = 2) is reading the temperature as 40degree and sends a short signal (which is not shown here) on

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.

Sensors Data Output

Example: Sensor Data Sent = 115

BS Data Received = 115

Sensors Data obtained at Base Station

Sensor IDs

Node ID = 2

Fig. 4. Simulation example of 15 Sensor Nodes with 8-bit data.

its turn (when ID_counter = 2), sharp at the beginning of level-5 of Data_Counter which is assigned a value of 40 (considering the offset value=35). On the other side, the same synchronized counters in the base station produce this value (40-degree) within the level-5 of the Data-Counter assigned to ID(2) which is obtained in ID_Counter right after detecting this short signal on the RX output of the receiver. Fig. 4 shows the same simulation results in a higher resolution. The ID_Counter is upgraded from 2-bit to 4-bit to identify up to 16 sensor nodes, and the Data_Counter from 3bit to 8-bit generating 256 levels through random number generators as input to all sensor nodes. As shown in this figure, the corresponding data to each sensor is obtained with slightly corresponding delay in each step. The simulation is done in larger scales as large as 8-bit for sensors and 10-bit per data (256 sensors with data range: 0, 1023). Since the sensor nodes use TDM technique to send data, this implies one sample per second for each sensor at the frequency of 256 KHz which is the baud rate. Obviously for higher network speed, say, s sample per second for each sensor, the frequency must be multiplied by s. Therefore, the clock frequency or the baud rate of data transaction could be easily obtained by (1). f = n . (2)d . s

(1)

where f is frequency of data transmission (baud rate), n is number of sensors, d is data width in bits and s is sample per unit of time by each sensor.

V. TIME SYNCHRONIZATION ANALYSIS Time synchronization in Wireless Sensor Networks (WSNs) plays an important role especially when dealing with big number of distributed sensors within the WSN structure. Indeed, time synchronization schemes or protocols developed for traditional networks such as NTPs [11] are ill-suited for WSNs and hence more appropriate approaches should be proposed. The most important characteristics that should be considered in time synchronization for WSNs are: lifetime, precision and efficiency. Actually none of the standard clock/time synchronization methods [8, 11] is energy-efficient for IWBSNs. They transact with very long messages through the sensor network and consequently keep the nodes in longterm listening mode. This will waste large amount of energy in the tiny energy sources associated with such distributed nodes. In this paper, a new time synchronization algorithm applicable to our proposed communication protocol for IWBSNs is introduced. Fig. 5 illustrates the transactions between three sensor nodes (S1, S2 and S3) and the BS in this algorithm. Whenever the BS detects a time drift from one or more sensor nodes, it stops reading the data from the sensors and enters the synchronization mode. Then it sends a short primary signal (Pr_Signal) continuously within one period of sensors cycle. This is to notifying all sensors to enter the synchronization mode in the subsequent sensor cycle. Sensor cycle is the time period in which all sensors send their messages once. In the subsequent sensor cycle (Sync_Cycle), the BS continuously sends a long message to all sensors. This message starts with a pre-code for the sensors to detect this message, followed by a Time-Stamp to assign the remaining time to reach synchronization instant.

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.

Sending Primary Signals

Sensor Components

Sends instantly Time-Stamps

BS Delay

Rx Tx

Battery

BS

Sensor

Amp.

ADC

Logic

u-Controller

Transceiver

Fig. 6. General parts of a sensor node. Ti

S1

Tk S2 Tj

S3

Sensors read Primary signal

Sensors read Time-Stamp

Synchronizing Instant

Fig. 5. Time Synchronization scheme.

instant. Obviously the Time-Stamps are different in each message, since they get smaller in subsequent messages. By default, each sensor is supposed to listen if there is a Pr_Signal from the base station at the beginning of each sensor cycle. Each time a sensor node detects a Pr_Signal, gets ready for transition to the long listening mode in the subsequent sensor cycle which is the Sync_Cycle. Every unsynchronized sensor node could enter the Sync_Cycle at any unknown time slot of this cycle, and then start seeking for the earliest arriving Time-Stamp message. Once the sensor receives the Time-Stamp, stores it in its Sync_Counter and starts decrementing it on each clock cycle. Once the Sync_Counter in each sensor node reaches zero, the sensor node resets its ID_Counter and Data_Counter. This is the event that all sensor nodes get reset and synchronized with the Base Station. In Fig. 5, the Time-Stamps which are read by the sensors S1, S2 and S3 are shown respectively as Ti, Tj and Tk. These Time-Stamps are calculated in the BS so that all decrementing Sync_Counters in all sensor nodes reach zero at the same time which is the synchronization event. The reason of using the primary period is to avoid the sensors operating always in long listening modes at the beginning of each sensor cycle. This way, the sensors will always be in the shortest possible listening mode until they receive a Pr_Signal, and then prepare themselves for the Sync_Cycle to read a Time-Stamp. VI. DISCUSSION & COMPARISON Irrespective of the synchronization phase which is a requirement in all types of sensor networks, the proposed protocol in the ideal case introduces a gain of up to 280 times

higher throughput than ZigBee protocol [2] in terms of the energy saving. Because the smallest packet in ZigBee consists of at least 280 bits of data to be sent per sample. Moreover, in such cases where the transceivers are based on FSK (Frequency Shift Keying) modulators or any other kind of PLL based modulators, there will be another problem with a long-term start-up time too. Then a significant amount of energy of the tiny batteries in the tiny nodes will be dissipated. However, in this work, one single transmitted data bit per round in the ideal case is used, and in order to alleviate the environment noise effect, data can be coded using only few bits instead, which is still a considerable difference versus the 280 bits minimum (composed of numerous control bits) in ZigBee protocol. Also, in this work there is no need to use the PLL-based modulators. ASK (Amplitude Shift Keying) modulators using OOK (On Off Keying) method is absolutely practical and applicable for this protocol. The performances of the new and fast emerged technology of the transceivers can guaranty quick transition from sleep mode to active mode especially in transmitter side. Fig. 6 shows a general structure of a sensor node consisting of six parts. The total power consumption of this node is:

PTotal = PAmp + PADC + PLogic + PuC + PTransceier

(2)

Among all the node parts, the transceiver part consumes the largest amount of energy versus the others. In the standard structure of a sensor node, a low-power microcontroller exists that can benefit from the latest technology of the Phoenix Processor (A 30pW Platform for Sensing Applications) [9] with ultra-low power (29.6pW) in sleep mode and 2.8 uW @ 1MHz in active mode. Also the available technology of the low power counters and comparators significantly alleviates the power issue of logic parts added in our new protocol. For example, a Real Time Clock chip like; “RTC DS-1372” which is a 64-bit counter with a power consumption of 400nA @ 3Volts (only 1.2uW) can be selected as our 16-bit counter. Current technology of the low-range transmitters like those used in ZigBee transceivers with the 0dBm output power setting which is applicable to WBSNs, consume about 100nA (0.33uW @3.3 Volts) in sleep mode and about 26mA or 85mW in active mode. Considering a duty cycle of about 0.1% (280/256K) in active/sleep mode using ZigBee standard communication protocol at 256 Kbps, the average consumption would be up to 81uW. Therefore, this is not

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.

comparable with the low energy dissipated within the added logic part, i.e. a 16-bit counter and a comparator, which is used in our proposed protocol. Also the ADC section (Fig. 6) consumes trivial amount of energy like in [10] with 31 pj/8bit sample @ 1 Volt which is extremely acceptable for body sensor sampling requirement.

REFERENCES [1]

F. Fereydouni_Forouzandeh, O. Ait Mohamed, M Sawan. “Ultra Low Energy Communication Protocol for Implantable Body Sensor Networks,” in Proc. IEEE NEWCAS-TAISA’08, Montreal, June 2008, pp. 57-60.

VII. CONCLUSION

[2]

CC2520 DATASHEET. 2.4 GHZ IEEE 802.15.4/ZIGBEE® RF TRANSCEIVER, SWRS068 – December 2007, Available: http://focus.ti.com.cn/cn/lit/ds/swrs068/swrs068.pdf

Implantable wireless body sensor networks (IWBSNs) have recently emerged as an important and growing research area. In this paper, we propose a novel protocol in which, instead of transmitting the value of the sampled data in each round, the data is coded in a time domain range and only a very short signal will be sent to the BS by the transmitter sharply at a corresponding level within this range. This will significantly extend the battery lifetime of the sensor node in human body by keeping the transceivers in their minimum activity status, which is the sleep mode, almost all the time. Hence our proposed utility protocol that is used for collecting, processing, and for communication purposes is designed not only to work reliably, but more importantly to be energyefficient. Also, a new time synchronization algorithm that is especially applicable to our proposed communication protocol is briefly introduced and is the focus of our future work.

[3]

Benny P L Lo, G. Z. Yang, “Key Technologies and current Implementations of Body Sensor Networks”, Imperial College London, UK, available: http://www.doc.ic.ac.uk/~benlo/ubimon/BSN.pdf.

[4]

V. Shankar, A. Natarajan, S. K. S. Gupta L. Schwiebert, „EnergyEfficient Protocols for Wireless Communication in Biosensor Networks”, Dept. of Computer Science and Engineering Computer Science Department Arizona State University Wayne State University Tempe, AZ 85287 Detroit, MI 48202.

[5]

Sun Limin, Li Jianzhong, Chen Yu, Wireless Sensor Networks, Tsinghua publishing company, Beijing, 2005.

[6]

W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energyeficient Communication Protocol for Wireless Sensor Networks,” in proceeding of the Hawaii International Conference on System Sciences, Hawaii, January 2000, 10 pp. vol.2.

[7]

Stephanie Lindsey Cauligi S. Raghavendra, “PEGASIS: Power-Efficient Gathering in Sensor Information Systems”, Computer Systems Research Department The Aerospace Corporation P.O. Box 92957 Los Angeles, CA 90009-2957.

[8]

Leslie Lamport, P. M. Melliar-Smith, “Synchronizing Clocks in the Presence of Faults”, Journal of the Assosiation for Computing Machinery, Vol.32, No. I, Jan, 1985, pp. 52-78.

[9]

ERC Center for Wireless Integrated MicroSystems (WIMS), University of Michigan, Michigan State and Michigan Technology University, Available: http://wimserc.org/thrusts/nuggets.php

ACKNOWLEDGMENT The authors would like to express their sincerest thanks to ReSMiQ (Le Regroupement Stratégique en Microsystèmes du Québec), an inter-university research center with the objective of establishing collaborative links with industrial and academic partners, for its great helps and supports. We also thank CMC (Canadian Microsystems Corporation) for the tools and support services provided for this project.

[10] M. D. Scott, B. E. Boser, K. S. J. Pister, “An Ultra-Low Power ADC for Distributed Sensor Networks”, ESSCIRC, 2002. [11] D. L. Mills. Internet Time Synchronization: The Network Time Protocol. In Z. Yang and T. A. Marsland, editors, Global States and Time in Distributed Systems. IEEE Computer Society Press, 1994.

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.