Related Work

As there are only few MAC-protocols for WBANs, we will start with the research in the area of wireless sensor networks. The two major categories of current MAC-protocols designed for wireless sensor networks are contention-based and schedule-based [1, 2]. For the former, CSMA/CA is a typical example, while TDMA is a typical scheme for the latter. The advantages of contention-based ap-proaches are the simplicity, its infrastructure-free ad hoc feature and good adapt-ability to traffic fluctuation, especially for low load. Schedule-based approaches on the other hand are free of idle listening, overhearing and packet collisions because of the lack of medium competition, but require tight time synchronization.

The most commonly used technique for reducing energy consumption in con-tention-based protocols is controlling the power and duty cycle of the radio.

The Sensor-MAC (S-MAC) protocol [3] uses scheduling to coordinate sleep-ing among neighborsleep-ing nodes to avoid idle listensleep-ing. A node tries to synchronize with its neighboring nodes. If no neighbor is found, it will choose a schedule to start with. These synchronized neighbors form “synchronized islands”. Neighbors in reach of two or more islands have to synchronize with all islands’ schedules.

An extension of the S-MAC is the T-MAC protocol [4], which allows the nodes to go back to sleep when no traffic is detected for a certain time. WiseMac [5]

and B-MAC [6] use a preamble sampling technique. Nodes wake up for a short period after intervals with fixed length. The wake-up period is at least as long as the preamble so whenever a node wakes up and it hears a preamble it knows it has to stay awake to receive packets after the preamble. Doing so, the receivers significantly reduce the energy used for idle listening. To reduce the overhead as-sociated with long preambles, a strobed sequence of short packets allowing for fast shutdown and response is used in [7]. An ultra-low duty cycle MAC that combines scheduling and channel polling is presented in [8]. All suffer from synchroniza-tion overhead and periodic exchange of sleeping schedules. D-MAC [9] can be summarized as an improved Slotted Aloha algorithm in which slots are assigned to the sets of nodes based on a data gathering tree. The principal aim is to achieve very low latency for converge cast communications, but still be energy efficient.

However, the set up of the tree is not flexible and collisions may occur when nodes on the same level try to send data.

One of the few MAC-protocols for WBANs was proposed by Lamprinos et al.[10]. They use a master-slave architecture and, to avoid idle listening, all the slaves are locked in the Rx-slot of the master and go in standby at the same time.

The main drawback of this protocol is that some slaves will have a low duty cycle whereas the nodes that are serviced later have a higher duty cycle. The protocol was implemented nor simulated. An adaptation of this protocol was used in [11].

This protocol divides time into frames in which only one node is allowed to trans-mit. The scheduling order is derived by applying the Earliest Deadline First

al-CHAPTER5 105

gorithm. Omeni et al. [12] propose a MAC protocol for a star-networked WBAN that supports TDMA to reduce the probability of collision and idle listening. Each slave node is assigned a slot by the central node. When an alarm occurs at one of the nodes, the node can be assigned an extra slot for direct communication. The protocol has been evaluated on a Sensium platform. The H-MAC protocol [13]

uses the human heartbeat rhythm information to perform time synchronization for TDMA. The bio-sensors can thus achieve time synchronization without having to turn on their radio. The algorithm is verified with real world data but assumes a certain buffer. The simulations do not show the energy gain and the protocol is designed for a star-topology WBAN only.

Most current implementations of WBANs use IEEE 802.15.4 [14] or Zig-Bee [15] as enabling technology. Some implementations use Bluetooth (IEEE 802.15.1) [16]. This was developed as a cable replacement and does not support (or only very limited) multi-hop communication. It has a complex protocol stack and a high energy consumption compared to IEEE 802.15.4. It is therefore not suited to be used in a WBAN. In the following section, we will discuss IEEE 802.15.4 and its appropriateness for communication in WBANs.

5.1.2 IEEE 802.15.4

The IEEE 802.15.4 standard [14] is designed as a low power and low data rate protocol offering high reliability. It defines the physical layer and the medium access layer and uses a beaconed and unbeaconed version. For the physical layer, 27 communication channels in three different frequency ranges are defined in the industrial scientific medical (ISM) band: 16 channels in the 2.4 GHz band, 10 channels at 915 MHz and 1 channel at 868 MHz. The 2.4 GHz band is available worldwide and operates at a raw data rate of 250 kbps. The channel of 868 MHz is specified for operation in Europe with a raw data rate of 20 kbps. For North America the 915 MHz band is used at a raw data rate of 40 kbps.

In [17] and [18] (added as Appendix A) we have analyzed both analytically and experimentally the maximum throughput and minimum delay of the unbeaconed or unslotted version of the protocol. The exact formula for the throughput and delay of a direct transmission between one sender and one receiver is given. This is done for the different frequency ranges and address structures used in IEEE 802.15.4. The analysis is limited to the unslotted version as this one experiences the lowest overhead. In section 5.1.2.1 we recapitulate the main results. For the full mathematical background and an overview of the IEEE 802.15.4 protocol, the reader is kindly referred to Appendix A.

106 NETWORKPROTOCOLS FORWBANS

5.1.2.1 Throughput and Delay

In the beaconless mode, a simple CSMA/CA protocol is used. When a device wishes to transmit data, the device waits for a random number of back off periods.

Subsequently, it checks if the medium is idle. If so, the data is transmitted, if not, the device backs off once again. The standard defines three types of addressing: 64 bit, 16 bit and no addresses. The acknowledgments can be omitted and the packet size is limited to 127 bytes at MAC-level, headers included.

0 20 40 60 80 100 120

0 2 4 6 8 10 12 14 16x 10⁴

Payload size (bytes)

Useful bitrate (bps)

address 16 bits + ACK address 16 bits no ACK address 64 bits + ACK address 64 bits no ACK

Figure 5.1: Useful bitrate for IEEE 802.15.4 in function of a varying payload size for the short and long address scheme, with and without ACK. The frequency is set to 2.4 GHz.

Figure 5.1 shows the maximum throughput for IEEE 802.15.4 in the 2.4 GHz region. The throughput is expressed as the useful number of bits, i.e. the data bits without the overhead or the MAC payload, that can be sent during 1 second. It is shown that the maximum throughput depends on the packet size. For smaller pack-ets the throughput is significantly lower as the ratio of the overhead increases. The throughput is also lower when acknowledgments are used, as expected. When the addressing scheme of 16 bits is used, the overhead is smaller and the throughput is higher. The payload size can also be larger as the packet size is limited to 127 bytes including headers. In the 2.4 GHz band, a bandwidth efficiency of 64.9%

is reached in optimal circumstances, i.e. when no addresses and no acknowledge-ments are used. If acknowledgeacknowledge-ments are used, an efficiency of merely 59.5% is obtained. Using the short address further lowers the maximum bit rate by about

CHAPTER5 107

4%. The worst result is an efficiency of only 49.8% which is reached when the long address is used with acknowledgements. The main reason for these low re-sults is that the length of the MPDU is limited to 127 bytes. Similar conclusion can be made in the other frequency bands, see Table A.3.

0 20 40 60 80 100 120

1 2 3 4 5 6 7x 10⁻³

Payload size (bytes)

delay (s)

address 16 bits + ACK address 16 bits no ACK address 64 bits + ACK address 64 bits no ACK

Figure 5.2: Minimum delay for IEEE 802.15.4 as a function of the payload size. The fre-quency is set to 2.4 GHz.

Figure 5.2 gives the minimum delay each packet experiences for varying packet sizes in the 2.4 GHz band. The minimum delay is calculated by sending a packet without any data bits immediately from 1 sender to 1 receiver. The propagation delay is not taken into account and no retransmissions are assumed. We immedi-ately notice that the delay is a linear function of the number of payload bytes, as long as we assume a payload of more than 10 bytes. The transmitted frames are followed by an Inter Frame Space (IFS) in order to allow the MAC layer a finite amount of time to process data received from the PHY. Before starting the back off period, the device will wait one IFS. Long frames (MPDU is larger than 18 bytes) are followed by a Long IFS (LIFS) and short frames by a Short IFS (SIFS). The jump in the graph for the short address length is caused by the this mechanism.

The same behavior is found for the other frequency bands.

The maximum delay is found by sending a full packet, i.e. the MPDU is set to the maximum of 127 bytes. The maximum delay is a little bit higher than 6 ms in the 2.4 GHz region when a full packet is sent, see Table A.4. This delay is acceptable for delay bound applications. The lower bands experience a significant higher delay, which is to be expected as the data rate is lower. In these frequency

108 NETWORKPROTOCOLS FORWBANS

bands it is more important to look to the minimum delay, especially in the 868 MHz band.

Further we have also investigated the influence of the back off interval. A significant gain is found when the back off exponent is lowered. For more infor-mation, see Appendix A.

5.1.2.2 IEEE 802.15.4 on the Body?

In [19] the star network configuration of the IEEE 802.15.4 standard at 2.4 GHz was considered for a WBAN. The analysis considers quite extensively a very low data rate star network with 10 body implanted sensors transmitting data 1 to 40 times per hour. The analysis focuses on the effect of crystal tolerance, frame size and the usage of IEEE 802.15.4 Guaranteed Time Slots (GTS) on a node lifetime.

The main consideration in this work was the long-term power consumption of devices. The results show that, even when properly configured, IEEE 802.15.4 provides a limited answer for medical sensor networking when configured in non-beacon mode with low data rate asymmetric traffic. Beacon mode may also be used, but with more severe restrictions on data rate and crystal tolerance.

Another adaptation is BSN-MAC [20]. The coordinator controls the communi-cation by varying the superframe structure of IEEE 802.15.4. This divides the time axis in a contention-free and contention-based period. The sensors provide real-time feedback to a BSN coordinator with application-specific and sensor-specific information. Hence, based on the feedback the BSN coordinator can make dy-namic adjustments for the length of the contention-free and contention-based pe-riod to achieve better performance in energy efficiency and latency.

Both [21] and [22] come to the conclusion that although 802.15.4 can provide QoS, the technology is not scalable in terms of power consumption and can not be used as a single solution for all WBAN applications. This view is shared by IEEE 802.15.6 that aims to develop a proper MAC for WBANs [23].

As such, it can be concluded that IEEE 802.15.4 is not the best solution for supporting communication in WBANs. Although it can be used for a quick (and easy) implementation, the results are rather poor. IEEE 802.15.4 was not designed to support WBANs. Specialized MAC protocols are needed. As a consequence, we will not use radios that implement the IEEE 802.15.4 PHY-layer, but more general low-power transceivers such as the Nordic nRF2404 transceiver (see Sec-tion 2.3.3.2).

5.2 Routing

A lot of research has already been performed in the area of wireless sensor net-works. An overview can be found in [24]. In the following, an overview of existing

CHAPTER5 109

routing strategies for WBANs is given.

When considering wireless transmission around and on the body, important issues are radiation absorption and heating effects on the human body. To reduce tissue heating the radio’s transmission power can be limited or traffic control al-gorithms can be used. In [25] rate control is used to reduce the bioeffects in a single-hop network. Another possibility is a protocol that balances the communi-cation over the sensor nodes. An example is the Thermal Aware Routing Algo-rithm (TARA) that routes data away from high temperature areas (hot spots) [26].

Packets are withdrawn from heated zones and rerouted through alternate paths.

TARA suffers from low network lifetime, a high ratio of dropped packets and does not take reliability into account. An improvement of TARA is Least Tempera-ture Routing (LTR) [27] that reduces unnecessary hops and loops by maintaining a list in the packet with the recently visited nodes. A combination of LTR and shortest path routing is Least Total Route Temperature (LTRT) [28]. The nodes’

temperatures are converted into graph weights and minimum temperature routes are obtained. A better energy efficiency and a lower temperature rise is obtained, but the protocol has as main disadvantage that a node needs to know the temper-ature of all nodes in the network. The overhead of obtaining this data was not investigated.

“Anybody” [29] is a data gathering protocol that uses clustering to reduce the number of direct transmissions to the remote base station. It is based on LEACH [30] that randomly selects a cluster head at regular time intervals in order to spread the energy dissipation. The cluster head aggregates all the data and sends it to the base station. LEACH assumes that all the nodes are within sending range of the base station. Anybody solves this problem by changing the cluster head selection and constructing a backbone network of the cluster heads. The energy efficiency is not thoroughly investigated and reliability is not considered. Another improvement of LEACH is Hybrid Indirect Transmissions (HIT) [31], which com-bines clustering with forming chains. Doing so, the energy efficiency is improved.

Reliability, however, is not considered.

Ruzelli et al. propose a cross-layer energy efficient multi-hop protocol built on IEEE 802.15.4 [32]. The network is divided into timezones where each timezone takes turn in the transmission. The nodes in the farthest timezone start the trans-mission. In the next slot, the farthest but one sends it data and so on until the sink is reached. The protocol almost doubles the lifetime compared to regular IEEE 802.15.4. The protocol was developed for regular sensor networks, but the authors claim its usefulness for WBANs.

This overview clearly shows that routing protocols for WBANs are an emerg-ing area of research. The protocols described above were only developed in the last two years. A combination of MAC-layer and routing layer in one protocol has not been considered yet. In the following, we will present a cross layer protocol,

110 NETWORKPROTOCOLS FORWBANS

WASP.