Prioritization of Patient Vital Signs Transmission Using Wireless Body Area Networks - E-Healthcare Systems and Wireless Communications - page 273

requirement for triage, where traffic from patients
in more critical conditions should be served earlier
than that from those patients in safer conditions.
To take the advantages of these two scheduling
schemes, we adopt a two-level Hybrid Scheduling
(HS) approach with the first being EDF scheduling
level to meet the e2e requirement of the flows and
the second being GPS scheduling level based on
the traffic class. Both levels adapt to the traffic
variations, thus maximizing the reliability of these
classes while meeting the flow e2e requirements.
As discussed in Sect. 2.3, hybrid schedulers such
as (Gidlund & Wang, 2009) and (Wongthavarawat
& Ganz, 2003) can apply either EDF or WFQ but
not both to a flow, which cannot satisfy the above
traffic requirements.
Consequently, as shown in Figure 5, two levels
of scheduling are employed to select a packet to
send. At the first level, flow-based scheduling ,
packets are placed into three classes (red, yellow,
and green) of queues according to the associ-
ated patient's condition. Then, EDF scheduling
is applied to choose a packet for each class in
the ascending order of the packets' estimated
expiration time. At the second level, class-based
scheduling , a GPS scheduler is proposed to select
one among the chosen packets at the first level for
transmission. Suppose the weights assigned to the
red, yellow, green classes of queues are ϕ R , ϕ Y ,
ϕ G , respectively, for the GPS scheduler, then it is
clear that we have ϕ R + ϕ Y + ϕ G =1 and ϕ R ≤ ϕ Y
≤ ϕ G based on the group's significance. Note that
these weights are adjusted adaptively according
to the feedback from the sink.
Compared to the Differentiated Services (Dif-
fServ) approach to provide QoS to traffic, as
defined in IETF RFC 2474, our hybrid scheduling
algorithm is different in several aspects: 1) our
algorithm can guarantee the e2e delay require-
ments of the vital signs, while DiffServ can only
serve the traffic based on the class requirements
and, thus, does not in general guarantee the delay
requirement of the vital signs; 2) our algorithm
is specifically tailored for patient emergency
status and delay requirements of vital signs, as in

Table 1, while DiffServ is designed for different
classes of traffic such as voice and video with
quite different requirements; 3) our algorithm can
adjust the traffic class coefficients adaptively
using feedback information from the sink, while
DiffServ generally does not use any feedback
mechanism and therefore is not able to adapt to
changes of network traffic. On the other hand,
compared with the Integrated Services (IETF RFC
1633) QoS approach, which treats each flow in-
dividually without bundling them into aggregates
belonging to the same class, our algorithm has
much lower complexity and better scalability
although a certain “bandwidth inefficiency” is
introduced, as will be discussed in Sect. 4.4.2.
In the following sections, we focus on the
feedback mechanism to maximize the reliability
of different classes of traffic with guaranteed e2e
delay requirements of different flows.

Class-Based Scheduling with Feedback

To maximize the reliability for these three classes
of traffic while guaranteeing e2e delay, an adaptive
feedback mechanism is used to adjust the GPS
weight triple (ϕ R , ϕ Y , ϕ G ) at each CH. Feedback
packet follows the reverse path from the sink to
the source. Depending on the feedback from the
sink, which includes the reliability value per class
and the average delay per flow, the parameters of
the scheduler (number of queues and fraction of
capacity to red/yellow/green queues) are adjusted
on each node along the path from the source to
the destination.
In our work, packets of higher class - red or
yellow - are delayed to reliably deliver traffic of
lower class - yellow and green -, respectively,
so that the reliability of the lower class is maxi-
mized. Note that the e2e delay requirements for
the higher classes are still guaranteed. A packet
is reliably delivered to the destination when it is
received before it expires, while expired packets
are dropped in the network no matter their class
(color).