• 沒有找到結果。

WBAN Quality of Services (QoS)

在文檔中 多人無線近身網路系統 (頁 62-0)

Chapter 4 Distributed Multiuser QoS Designs

4.1 WBAN Quality of Services (QoS)

WBAN QoS controls that simultaneously support both intra and inter WBAN QoS are studied in this work. A WBAN consists of a single central processing node (CPN) and several wireless sensor nodes (WSNs). These WSNs collect various medical data (including vital signals from human body and diagnosis audio/video) and forward them to the CPN, which is depicted in Fig. 4-1. Intra WBAN QoS controls should make sure these medical data are timely transmitted by following their delay-bound and delay-variation requirements [51]. However, when total bandwidth requirements of a WBAN overflow its capacity, transmissions should be scheduled in an order from the highest-priority data to that has the lowest priority, which guarantees the QoS level of high priority data. Such priority settings are usually designed by medical experts according to their clinical experiences. For example, a heart failure could introduce much instant life risk than an abnormal body-temperature. Hence, ECG signals directly reflecting heart activity should have higher priority than that of temperature records. This kind of priority is called as intrinsic data priority.

Furthermore, if abnormal vital signals are detected, the priorities of these signals should be

51

dynamically increased to be higher than those of normal signals. Such priority is called as emergent data priority. Therefore, intra WBAN QoS controls should meet various latency requirements of difference medical data and follow proper intrinsic and emergent data priorities simultaneously.

On the other hand, for inter-WBAN QoS designs, proper user priority should be further provided.

In scenarios of multiple overlapped WBANs, WBANs need to share radio resource with each other.

Once the overall capacity is not sufficient to support all transmission bandwidth of WBANs, radio resources should be allocated to WBANs that has higher user priority. Such priority should also be defined by medical experts. Usually, priority settings follow an order from the highest to the lowest life-critical WBAN users. High user-priority WBANs should be allowed to transmit all necessary medical data; low user-priority WBANs should transmit only partial medical data to maintain normal health monitoring. As a result, WBAN QoS controls should simultaneously satisfies (i) intrinsic data priority (ii) emergent data priority and (iii) user priority for both intra and inter WBAN QoS.

Aside from transmission qualities above, a WBAN QoS control should try to lower energy consumption of WSNs as well [1, 51, 52]. In a WBAN, a CPN will most likely be embedded in personal devices such as cellular phones or PDAs with larger and rechargeable batteries. In contrast, WSNs are expected to be light weight (small battery) and even un-rechargeable for certain implantable applications. Thus, WSNs are expected to keep their energy consumptions as low as possible.

4.1.B Performance Metrics

To qualify a QoS control for WBAN, following performance metrics will be evaluated.

52

 Transmission Latency: transmission latency affects smoothness of real-time display of vital signals. A transmission latency of a medical packet is calculated from the time of a packet is generated in a WSN to the time of the packet is successively received by a CPN. To ensure a vital signal is timely displayed, every packet of the signal should be received before its delay bound expires.

 Joule per bit of WSN: energy consumption of a WSN affects its battery life. To evaluate energy consumptions of WSNs with various traffic loading, an energy measurement is normalized by its transmission bandwidth with the unit, Joule per bit. An energy measurement of a WSN will count all its packet transmissions (successful/unsuccessful packet transmissions from WSN to CPN) and receptions (successful/unsuccessful polling message receptions from CPN to WSN).

 User capacity: user capacity affects the density of coexistence WBAN users, which is important for dense WBAN scenarios. User capacity is defined as the maximum number of coexistence WBANs that satisfy desired WBAN QoS requirements.

4.1.C Related Works

Significant contributions toward high quality WBAN QoS designs have been made in recent years [36-41, 53, 54]. These works adopt different framing, scheduling, and novel hardware techniques to optimize emergency transmission, packet latency, and power consumption of a single user WBAN.

Huasong [38] creates a framing-structure-turning procedure to simultaneously improve throughput, queuing delay, and energy consumption of IEEE 802.15.4, a candidate protocol for WBAN. Yoon [36]

further modifies the framing structure of 802.15.4 to remarkably reduce the packet delay of

53

emergency alarm. He further introduces a preemptive scheduling to guarantee the transmission priorities of various medical data. There are also scheduling techniques utilizing TDMA-overhead-reduction [53], adaptive duty cycle [54], prioritized retransmission [39], delayed retransmission [37], fuzzy-logic controls [41], and wake-up radio [54] to enhance WBAN QoS. Su and Zhang further combine scheduling with realistic battery charging/discharging effect to significantly prolong battery life of WBAN sensors. More complete introductions and comparisons of existing WBAN QoS solutions are summarized by Ullah [52]. Different from above single WBAN solutions, proposed RACOON protocol puts more focus on multi-user WBAN QoS solution, which will be introduced in following sections.

4.2 Random Contention-based Resource Allocation (RACOON)

The proposed Random Contention-based Resource Allocation (RACOON) is a bandwidth control system embedded in medium access control (MAC) layer for multi-WBAN QoS, which consists of two major designs: a CPN-based resource allocation and a random contention-based inter-CPN negotiation.

4.2.A CPN-based Resource Allocation

The CPN-based resource allocation of RACOON is designed to minimize energy consumptions of WSNs and early detect inter-WBAN interference to avoid unnecessary packet collisions. The proposed CPN-based protocol has a two-step resource allocation scheme, which is illustrated in Fig.

4-2. There are two distinct channels for inter and intra-WBAN communication, respectively. The inter-WBAN channel is used to exchange the resource negotiation messages between WBANs. Only CPNs can access the inter-WBAN channel. On the other hand, the intra-WBAN channel is used to transmit polling messages from CPN to WSN and data packets from WSN to CPN. The superframe

54

is divided into fixed number of slots. Each slot is sub-divided into a short polling slot and a data slot, which are illustrated in Fig. 4-2. WSN receive not only transmission schedule from polling messages but also framing structure information including the start of a superframe and number of slots in it.

When multiple WBANs overlap with each other, a CPN first negotiates WBAN resources with adjacent CPNs through the inter-WBAN channel. The detailed procedure of inter-CPN negotiation will be introduced in section 4.2.B. The CPN then assigns reserved resources to its WSNs by polling messages through the intra-WBAN channel. As a result, the WSNs wake up only when (1) receiving polling messages from the associated CPN in polling slots and (2) transmitting vital signals to that CPN if they are polled, hence energy consumptions of the WSNs can be reduced.

WBAN 2

Fig. 4-2 CPN-based resource allocation of RACOON

In the proposed CPN-based resource allocation, WSNs do not need to perform any interference detection. Instead, a probing-based interference detection, which utilizes a coverage difference between CPN and WSNs, is used and illustrated in Fig. 4-3. A CPN has a larger transmission range than that of WSNs, which allow the CPN to detect potential interferences to its uplink (WSN to CPN)

55

and downlink (CPN to WSN) transmissions. The detection is realized by periodic “probing” from the CPN to its adjacent CPNs. These probing messages are exactly the inter-CPN negotiation messages and the CPN is probed when it receives negotiation messages from other CPNs. Therefore, the interference detection and inter-WBAN negotiation are finished at the same time. Fig. 4-3 illustrates the proper range settings of a CPN and a WSN. For example, in an uplink (WSN to CPN) transmission with WSN’ as the source of interference (Soi), which is shown as case (a) in Fig. 4-3, CPN (CPN’) and WSN (WSN’) have the transmission ranges RCPN, and RWSN, respectively. For simplicity, the range of possible WSN position is assumed as a cylinder and shadowing effects of human body are ignored. CPN is located at the center of cylinder and WSNs are located within the cylinder. In case (a), a data packet is transmitted from WSN to CPN. In the mean time, WSN’ is transmitting data as well. To avoid a data collision happens at CPN, the distance from the interference edge of WSN’ (Soi) to CPN (Rx), that is, DEsoi Rx2W C' RWSN

should be positive.

Because the location of WSN’ is confined by the cylinder, the minimum W C'

lies on the C C'

connection. For this reason, the minimum RCPN that makes W C' RWSN

positive should be larger

than

2

WSN d

R  , where d is the diameter of the cylinder. CPN thus can detect the neighbor WBAN from radio activities of CPN’ before the interference of case (a) happens. Results of other combination of possible transmission directions and sources of interference are also presented in Fig.

4-3. By considering all cases, RCPN should be at least larger than RWSNd to ensure collision free transmissions.

56

' WSN

W C R

Tx→Rx Soi DEsoi2Rx Condition that CPN detects interf.

from Soi with Cylinder Model

Fig. 4-3 Probing-based interference detection

4.2.B Random Contention-based Inter-CPN negotiation

The inter-CPN negotiation of RACOON is an iterative bandwidth control scheme adjusted by two parameters: Bandwidth Requirement and User Priority Index. These two parameters are calculated by each CPN according to the status of it associated WSNs. From Fig. 4-4, each WBAN iteratively contends wireless resources to achieve a pre-defined bandwidth target. Besides, to reflect the emergency level (user priority index) of WBAN, two trends of bandwidth control scheme are provided for high and low priority WBANs, respectively. High priority WBANs aggressively contend resources to achieve a high bandwidth target, BWDesire. Bandwidth requirements of a low priority WBANs are relaxed to be in between BWDesire and a lower target BWRe quire. Thus, when

57

high and low priority WBANs contend to each other, high priority WBANs are expected to achieve better bandwidth targets than those of low priority WBANs can achieve.

Priority (SP/DP)

Trend of BW control for Low Priority WBAN RACOON

BWDesire BWRequire

High Priority Low Priority

Trend of BW control for High Priority WBAN

RACOON Input

Iterative BW control of inter-CPN negotiation

Fig. 4-4 Random Contention-based Inter-CPN negotiation of RACOON

The calculation of the bandwidth target, BWDesire, BWRe quire, and User Priority Index depends on proposed priority system for WSN. Each WSN has two kinds of priority indexes: static priority (SP) and dynamic priority (DP). Both SP and DP are set to either 0 (low) or 1 (high) depending on its

58 associated sensors of a WBAN user.

In RACOON, with the inputs of BWDesire, BWRe quire, and User Priority Index, each CPN generates a weighted random value to contend resources with its neighbor CPNs. The scheme of weighted random value is inspired by Neighborhood-aware Contention Resolution (NCR) algorithm [55], which provides collision free scheduling. The skill of NCR is a random value comparison scheme. Each wireless node first generates a random value. The wireless node that has the largest random value wins the transmission slot. As for the weighted random value contention in our case, it can be realized by a pseudo contention. A CPN first generates its Nrnd uniform random values and picks the largest one for a slot contention. The average probability P that CPN i can obtain a slot i is proportional to the number of random values,

rndi

N , used in the pseudo contention. That is

, ( )

Nrnd to achieve their bandwidth targets according to the control flow illustrated in Fig. 4-5. At the start of each contention iteration, the available bandwidth BWAvb, which is defined as the bandwidth

59

that a WBAN has in its previous iteration, is compared with its two bandwidth targets, BWDesire and

Re quire

BW . Then a three-case decision is decided basing on that:

Case 1: BWAvbBWRequire

Case 2: BWRequireBWAvbBWDesire

Case 3: BWDesireBWAvb

The value of Nrnd will be changed based on the priority setting of a WBAN. As shown in Fig.

4-5, high-priority WBANs are designed to increase their Nrnd more aggressively than those of low-priority WBANs. Thus, the high priority WBANs are more possible to achieve their bandwidth requirements than the low priority WBANs. In the proposed design, an iteration of contention is performed in a superframe. A CPN first generates random values for all slots according to Nrnd respectively and broadcasts only one negotiation message carrying these values through the inter-WBAN channel to its adjacent CPNs. Thus, contentions can be performed by value comparisons between CPNs. To avoid collisions between negotiation-messages, these messages are broadcasted at random time slots within a superframe.

60

Case2: BWRequire<=BWAvb<=BWDesire Nrnd: #Random Values slots according to Nrndand broadcasts negotiation message carrying these

Fig. 4-5 Bandwidth control flow of inter-CPN negotiation in RACOON

After a CPN finish an inter-WBAN resource contention, it performs an intra-WBAN scheduling to allocate its reserved transmission slots to its WSNs. A multi-queue scheduler is adopted to schedule resource by following order from the WSN that has the highest value of SP+DP to that has the lowest value, which is depicted in Fig. 4-6. A data in a queue is scheduled only when there is no data in other queues that has higher SP+DP value.

61

Fig. 4-6 Multi-queue scheduler of intra-WBAN scheduling

4.3 Computer Simulation

4.3.A Benchmarking WBAN QoS Protocol

BodyQoS [42] is a MAC layer scheduling scheme chosen to benchmark the proposed WBAN QoS protocol. BodyQoS is a CPN-centralized single-WBAN-QoS control that is capable of overcoming performance impacts from co-channel interference. The design strategy of BodyQoS is to increase transmission opportunities of a WSN when it suffers a bad channel condition. The transmission opportunity is inversely proportional to its available bandwidth, hence a vital signal can be timely transmitted without extra delay. The bandwidth control of BodyQoS can be expressed as:

( )

( 1)

ideal

ideal Avb

TxOpportunities t BW

TxOpportunitiesBW t

( 1) ( 1) (1 ) ( 2)

Avb Measured Avb

BW t   BW t   BW t

(4-2)

where BWideal is the ideal bandwidth with perfect channel; BWAvbis calculated by a moving average of previous measured bandwidth BWMeasured. To fairly compare the BodyQoS with the

62

proposed RACOON protocol, bandwidth measurement of the BodyQoS is performed every superframe. Besides, to avoid unpredictable interference, transmission opportunities are randomly scheduled within every superframe.10

4.3.B Experimental Settings

A MATLAB simulation platform is built to evaluate the proposed RACOON protocol. Detail settings of topology, PHY radio, MAC framing, and traffic loads, are listed in table I. Besides, to simulate WBAN mobility, the location change of WBANs follows the Gauss-Markov mobility model [33]. We use [33] to simulate the smooth movement path of a human, while avoiding the sudden stops and sharp turns that happen in the random walk mobility model [34]. The Gauss-Markov mobility model has a tuning factor  to control the randomness of WBAN movement.  is set as 0.5 in this study (0and 1 direct to a Brownian motion and linear movement respectively).

10 The original interference-avoidance scheme of BodyQoS is a Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) protocol. A random scheduling is used to simulate the random backoff skill of CSMA/CA.

63

TABLE 1 Experimental Setting of WBAN QoS

Item Value

Framing structure 10 slots per superframe / each slot is 50ms long

Packet size 240bytes

Transmission rate 48kbps

RX Power 27.3mW

TX Power 31.2mW

Topology 1 to 10 WBANs randomly deployed in a 6x6m2square. When number of WBANs is more than one, the ratio between high and low priority WBANs is 1:1.

Transmission range Distance between CPN and WSN: 0.5m. Referring to the range settings of probing-base interference detection in section 4.2.A, CPN: 3m / WSN: 2m.

Data Type Data Rate Data Priority (SP, DP) Delay Bound

ECG ch1 4kbps 1, 1 1s

ECG ch2 4kbps 1, 1 1s

SpO2 3kbps 1, 0 1s

Blood pressure 3kbps 0, 1 10s

Temperature 2kbps 0, 0 10s

Heart rate 2kbps 0, 0 10s

4.3.C Experimental Results

Packet latency of different vital signals in Fig. 4-7 illustrate how intra and inter WBAN priorities are realized in RACOON. For either high or low priority, latency of vital signals are ranked in order of SP+DP. Signal with higher SD+DP value should have lower latency. Note that SP and DP reflect the intrinsic and emergent data priorities, respectively. Furthermore, due to that the high-priority WBAN contends resources more aggressively than the low-priority WBAN does, same signal with same priority setting in the high-priority WBAN has shorter latency than that in the low-priority WBAN. This meets the QoS requirements of the user priority.

64

Fig. 4-7 Packet latency of vital signals in high and low priority WBANs

The latency comparison between RACOON and BodyQoS [42] in mobile WBAN scenarios is shown in Fig. 4-8. RACOON has much lower packet latency than BodyQoS has when WBANs move at either 2m/s or 6m/s. The reason is that RACOON makes WBANs cooperatively share the radio resource when they overlap to each other. On the contrary, BodyQoS does not consider interference interactions between WBANs, which makes improper decisions of bandwidth control and thus induces high transmission delay. In the original ideas of BodyQoS, interference is assumed to be generated by regular co-channel communications or path-loss due to limb movements. These sources of interference have “passive” interference patterns, which means it does not increase or

65

decrease its interference level following the bandwidth control of BodyQoS. Therefore, BodyQoS reasonably increases transmission opportunities to overcome bad channel conditions. However, in multi-WBAN scenarios, the increasing transmission opportunities cause serious inter-WBAN interference. It than increases the transmission opportunities again and causes more serious interference, which enters a vicious circle. Collision measurements with RACOON and BodyQoS in Fig. 4-9 echo this observation.

1 2 3 4 5 6 7 8 9 10

100 101 102 103 104

Number of WBANs

Packet Latency(slots)

BodyQos ECG 2m/s BodyQoS ECG 6m/s Racoon ECG 2m/s Racoon ECG 6m/s

Fig. 4-8 Packet Latency with RACOON and BodyQoS

66

1 2 3 4 5 6 7 8 9 10

0 100 200 300 400 500 600 700 800 900 1000

Number of WBANs

Packet Collisions (times)

BodyQoS 2m/s BodyQoS 6m/s Racoon 2m/s Racoon 6m/s

Fig. 4-9 Packet collisions with RACOON and BodyQoS controls

The collision measurements also show that collision of WBAN with RACOON is less sensitive to the number of co-existence WBANs, as compared with BodyQoS. Interference between WBANs is overcome by RACOON’s cooperative inter-WBAN resource sharing scheme and proposed probing-based interference detection. Collisions of RACOON are created by out-of-date scheduling.

A WBAN moves and encounters other un-negotiated WBANs with out-of-date inter-WBAN scheduling and thus packet collisions happen. However, besides of collisions created by WBAN mobility, BodyQoS creates extra collision by its problem of the inter-WBAN-interference enhancement. This problem gets worse when number of co-existence WBANs increases and hence

67

introduces more collisions. The difference reasons of collisions of RACOON and BodyQoS are also reflected in their energy consumptions, which is depicted in Fig. 4-10. Note that the energy consumption is normalized to transmission throughput. The energy consumption considers both TX and RX according to the definition in performance metrics, section 4.1.B.

1 2 3 4 5 6 7 8 9 10

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8

4x 10-3

Number of WBANs

Power Consumption (Joule per bit)

BodyQoS ECG 2m/s BodyQoS ECG 6m/s Racoon ECG 2m/s Racoon ECG 6m/s

Fig. 4-10 Energy consumption of WSN with RACOON and BodyQos

There is an interesting result in Fig. 4-8, 4-9, and 4-10. While mobility of WBAN user is increased, BodyQoS and proposed RACOON have opposite reactions. For BodyQoS, the latency, collision, and energy consumption of a WBAN are decreased. On the contrary, for RACOON, those of a WBAN are increased, which are shown in Fig. 4-8, 4-9, and 4-10. The reason comes from the

68

different anti-interference strategies of them. As for BodyQoS, a WBAN increases its transmission opportunities when it suffers inter-WBAN interference. If a collision history of a WBAN is separated

different anti-interference strategies of them. As for BodyQoS, a WBAN increases its transmission opportunities when it suffers inter-WBAN interference. If a collision history of a WBAN is separated

在文檔中 多人無線近身網路系統 (頁 62-0)

相關文件