mac scheduling in industrial wireless cell-based mesh
TRANSCRIPT
MAC SCHEDULING IN INDUSTRIAL
WIRELESS CELL-BASED MESH SENSOR
NETWORKS
Imran Yousaf
Master’s Degree Thesis 2011
EMBEDDED SYSTEMS
Postadress: Besöksadress: Telefon: Box 1026 Gjuterigatan 5 036-10 10 00 (vx) 551 11 Jönköping
MAC SCHEDULING IN INDUSTRIAL
WIRELESS CELL-BASED MESH SENSOR
NETWORKS
Imran Yousaf
This thesis work has been carried out at the School of Engineering, Jönköping
University within the subject area of embedded systems. The work is a part of
the Master programme in Electrical Engineering. The author takes full responsibility for opinions, conclusions and findings
presented.
Examiner: Youzhi Xu
Supervisor: Youzhi Xu
Scope: 30 credits (second cycle)
Date: December 07, 2011
Abstract
iii
Abstract
Undergoing developments for the adaptation of Wireless Sensor Networks (WSNs) in
automation industry is creating several research questions. The usage of wireless
technologies in industrial environments requires real-time reliability and security.
Many standards for WSNs and Industrial Wireless Sensor Networks (IWSNs) have
been developed. The list includes WirelessHART [4], ISA100.11a [10], Zigbee [16],
IEEE 802.15.4, and IEEE 802.15.4a etc.
Recently a new network topology “Cell-Based Mesh Networks” [12] has been
developed in an effort to make IWSNs feasible for large scale deployments in process
automation industries. Cell-Based Mesh Network topology inherits the qualities of
star-mesh topology and mesh topology, and also offers many additional features.
Reliability of an IWSN significantly depends on the real-time scheduling of the
network. During this thesis work a scheduling mechanism has been developed by
exploiting the features of a Cell-based Mesh Network. The idea of superframes’
further partitioning into time-cells of 100ms duration is presented, these 100ms time-
cells guarantees interlocking data transmission [9] within 100ms. The closed loop
control data [9] transmission is guaranteed by assigning dedicated time-slots. The
performance evaluation of the presented scheduling mechanism is done by
considering two case studies i.e. oil-industry and paper & pulp industry.
Keywords
iv
Keywords
Wireless Sensor Networks (WSN)
Industrial Wireless Sensor Networks (IWSN)
Wireless Sensor and Actuator Networks (WSAN)
WirelessHART
Cell-Based Mesh Networks
MAC Scheduling Algorithms
Real-time scheduling for IWSN
Network Topologies for WSNs
MAC Scheduling in WirelessHART
Acknowledgements
v
Acknowledgements
First of all, I would like to thank Allah almighty for all his blessings that he bestows
on us.
I would like to thank my supervisor Prof. Youzhi Xu, he has been very encouraging
and helpful for me during the whole thesis work. I really liked the way he explains the
research problems and involves the student in such a way that he can easily come up
with new ideas and innovations.
I would also like to thank our programme coordinator Alf Johansson for his guidance
and encouragement throughout the master programme and to all my teachers who
have given me invaluable knowledge during this master programme and to all my
colleagues.
Finally, I would like to thank my parents and family for their prayers, support, and
love throughout my whole life.
Contents
vi
Contents
Abstract ......................................................................................... iii
Keywords ....................................................................................... iv
Acknowledgements ........................................................................ v
Contents ........................................................................................ vi
List of Abbreviations ................................................................... viii
1 Introduction ............................................................................. 1
1.1 WIRELESS SENSOR NETWORKS .............................................................................................. 1 1.2 INDUSTRIAL WIRELESS SENSOR NETWORKS ........................................................................... 2 1.3 THESIS OBJECTIVES AND TASKS ............................................................................................. 3 1.4 THESIS STRUCTURE ............................................................................................................... 3
2 Background .............................................................................. 4
2.1 CELL-BASED MESH NETWORKS ............................................................................................. 4 2.2 WIRELESSHART PROTOCOL .................................................................................................. 5 2.3 SCHEDULING IN WIRELESSHART......................................................................................... 10
2.3.1 Location of Scheduling Algorithm in WirelessHART ................................................... 10 2.3.2 Requirements for Scheduling ...................................................................................... 10 2.3.3 Performance Measurers for MAC Scheduling Algorithms ........................................... 11
2.4 STATE-OF-THE-ART SCHEDULING MECHANISMS .................................................................. 11 2.4.1 Optimal Branch-and-Bound (B&B) Scheduling .......................................................... 11 2.4.2 Conflict-aware Least Laxity First (C-LLF) Scheduling ............................................... 12 2.4.3 Communication Schedule Defined in [1] .................................................................... 12
3 Design and Implementation ................................................. 14
3.1 PROBLEM FORMULATION ..................................................................................................... 14 3.1.1 Payload ..................................................................................................................... 14 Monitoring and Supervision ..................................................................................................... 14 Closed Loop Control ............................................................................................................... 14 Interlocking ............................................................................................................................. 15 Network Management .............................................................................................................. 15 3.1.2 Network Topology ...................................................................................................... 15 3.1.3 Superframe ................................................................................................................ 15 3.1.4 Scheduling ................................................................................................................. 16 Time-Slots ............................................................................................................................... 16 Dedicated ................................................................................................................................ 16 Shared ..................................................................................................................................... 16 Reserved.................................................................................................................................. 16 Unused .................................................................................................................................... 16
3.2 SCHEDULING ALGORITHM .................................................................................................... 16 3.2.1 Flow Diagram ........................................................................................................... 19
3.3 IMPLEMENTATION................................................................................................................ 21
4 Performance Evaluation ........................................................ 24
4.1 CASE-1: OIL INDUSTRY ........................................................................................................ 24 4.1.1 Schedule for Cell-1 .................................................................................................... 27 4.1.2 Schedule for Cell-2 .................................................................................................... 29
4.2 CASE-2: PAPER AND PULP INDUSTRY .................................................................................... 32 4.2.1 Schedule for Cell-1 .................................................................................................... 35 4.2.2 Schedule for Cell-2 .................................................................................................... 37 4.2.3 Schedule for Cell-3 .................................................................................................... 39
Contents
vii
5 Conclusions and Future Work .............................................. 41
5.1 CONCLUSION ....................................................................................................................... 41 5.2 FUTURE WORK .................................................................................................................... 41
6 References .............................................................................. 42
7 Appendices ............................................................................. 44
7.1 APPENDIX A STATE-OF-THE-ART NETWORK TOPOLOGIES ..................................................... 44 7.1.1 Star Topology ............................................................................................................ 44 7.1.2 Mesh Topology .......................................................................................................... 44 7.1.3 Star-Mesh Topology ................................................................................................... 44 7.1.4 Hub-and-Spoke Topology ........................................................................................... 45
7.2 APPENDIX B SCHEDULE TABLES OF CASE-2 .......................................................................... 45
List of Abbreviations
viii
List of Abbreviations
WSN Wireless Sensor Network
IWSN Industrial Wireless Sensor Network
HART Highway Addressable Remote Transducer Protocol
DLL Data Link Layer
B&B Optimal Branch-and-Bound Scheduling
C-LLF Conflict-aware Least Laxity First
TDMA Time Division Multiple Access
MAC Media Access Control
ASN Absolution Slot Number
DSSS Direct Sequence Spread Spectrum
ISM Industrial, Scientific, and Medical radio bands
OSI Open Systems Interconnection
Introduction
1
1 Introduction
The developments in wireless communication technologies have proven to be part of
various aspects of automation that include buildings, home, shipboard, transportation
systems, industry etc. Every environment has various distinct requirements for the
deployment of a wireless communication system e.g. for a temperature control system
in a building, the coverage area should be large in contrast to an industrial process
control system where the timing requirements are stringent though coverage area can
also be large in this case but not so important. Such wireless communication has been
made possible with the help of wireless sensor networks (WSNs). Research and
development at a large scale is undergoing to take this evolutionary step in the area of
automation. Many WSN standards have been developed though there are still lots of
research problems that need to be solved since WSNs are targeting a wide range of
environments.
This thesis work is related to industrial WSNs, that has stringent requirements for
wireless communication. Recently a new network topology called ‘Cell-Based Mesh
Networks’ [12] has been developed for industrial WSNs. In this thesis report a mac
layer scheduling mechanism for Industrial Wireless Cell-Based Mesh Networks has
been proposed.
1.1 Wireless Sensor Networks
A wireless sensor network (WSN) consists of spatially distributed autonomous
sensors to monitor physical or environmental conditions, such as temperature, sound,
vibration, pressure, motion or pollutants and to cooperatively pass their data through
the network to a main location [8]. Each of the sensor nodes of a WSN is equipped
with a radio transceiver, a small microprocessor and a number of sensors. The sensor
nodes are capable of forming the network autonomously and wirelessly communicate
with a central control, the data transmission can be in either direction. An abstract
picture of a WSN is shown in Fig 1.1a.
Figure 1.1a A Wireless Sensor Network
Introduction
2
Wireless sensor networks are bringing a revolutionary change in a wide variety of
application areas. These areas include;
Environmental monitoring
Healthcare
Critical industrial areas
Warehouse and supply chain
Military surveillance
Building automation etc.
Let’s take the example of environmental monitoring; a WSN can be deployed in a
forest that has caught fire, to monitor the further consequences and temperature in the
surroundings. The deployment can be done via an airplane since the sensor nodes are
autonomous so these will create a network autonomously and transmit data to the
central control to help the fire fighters to cope with the situation in a more efficient
way.
1.2 Industrial Wireless Sensor Networks
An industrial wireless sensor network can be defined as a WSN that can fulfil
stringent requirements on security and real-time reliability. Mostly the IWSNs are
centrally controlled and the dominate data traffic is sensed data from sensors to the
central control and actuation data to the actuators from the central control. A typical
industrial wireless sensor network is depicted in Fig 1.2a.
Figure 1.2a A typical Industrial Wireless Sensor Network [4]
These days, developments are on the go to make large scale deployment of wireless
communication technologies possible in industrial environments. Recently new
Introduction
3
standards have been developed that are particularly targeting industrial automation
domain, these include WirelessHART [4], ISA 100.11a [10] etc.
1.3 Thesis Objectives and Tasks
The main objective of this thesis project is to design and implement a feasible
scheduling algorithm/mechanism for Industrial Wireless Cell-Based Mesh Networks.
Although, the target IWSNs can be based on WirelessHART, ISA-100 or other
similar standards based on IEEE 802.15.4 but in this thesis work we will stick to the
WirelessHART standard for design and implementation details. More precisely
WirelessHART standard will be leveraged by cell-based mesh network topology for
network routing and for communication scheduling a feasible mechanism will be
proposed. Performance evaluation of the proposed scheduling algorithm will be for
one cell of a cell-based mesh network, the complete network’s performance
evaluation is out of scope of this thesis work.
1.4 Thesis Structure
The remainder of this thesis has been structured as follows. In Chapter 2, the
significance of scheduling in WSNs will be described by introducing Cell-Based
Mesh Networks and taking the case of WirelessHART protocol for scheduling. The
design and implementation of the proposed scheduling algorithm is described in
Chapter 3 and the performance evaluation is done in Chapter 4. The thesis report is
concluded in Chapter 5, some possible future work related to this project is also given
there.
Background
4
2 Background
This chapter is meant for describing the significance of scheduling in WSNs. Firstly,
Cell-Based Mesh Networks and WirelessHART protocol are introduced, and then the
significance of communication schedule in WirelessHART and State-Of-The-Art
Scheduling Mechanisms are given in this chapter.
2.1 Cell-Based Mesh Networks
State-of-the-art network topologies1 for IWSNs include star topology, mesh topology,
star-mesh topology, and hub-and-spoke topology, but IWSNs require a more suitable
network topology that can fulfil the stringent requirements of IWSNs that include
security, reliability, real-time, and low power consumption. Mostly the IWSNs are
centrally controlled and the dominate data traffic is sensed data from sensors to the
central control and actuation to the actuators from the central control. Cell-based mesh
network topology has been developed to provide a feasible solution for IWSNs.
Figure 2.1a A Cell-Based Mesh Network
The basic architecture of a Cell-based mesh network is depicted in Fig 2.1a. Its basic
elements include field devices, backbone routers, a gateway, a security manager, and
a network manager. The network is built up by linking multiple cells or clusters via
backbone to the network manager. Each cell has a backbone router that works as
cluster head, responsible for providing basic links to each field device with few hops
one or two and maximally 3 hops in the cell. Redundant routes are also available in
the form of intra-cell and inter-cell links [12]. Two frequency channels will be utilized
by each cell.
By inheriting the properties of a multiple-star-clusters network and a mesh network, a
cell-based mesh network provides a solution with significant improvements. The main
advantageous properties [12] are as follows:
i. High network reliability 1State-of-the-art network topologies are described in Appendix A.
Background
5
ii. Reduced retry times
iii. Improved network size and scalability
iv. Low latency
v. Simplicity
vi. High time synchronization accuracy
vii. High failure tolerance
2.2 WirelessHART Protocol
WirelessHART protocol is part of the HART 7 (www.hartcomm.org) standard. HART
was developed in the late 80’s. Initially HART communications protocol offered a
simple two-communications by using 4-20mA signals. After more than two decades
HART has evolved from its simple two-communication capability to the wired and
wireless communication capabilities with extensive features that includes security,
unsolicited data transfers, event notifications, block mode transfers, and advanced
diagnostics. The evolution of HART standard is depicted in Fig 2.2a. The most recent
version of HART standard is version 7 or simply HART 7. Along with many other
new features HART 7 introduced wireless mesh networking in the form of
WirelessHART communication protocol.
Figure 2.2a Evolution of HART standard [3]
WirelessHARTTM
protocol is designed for communication within Wireless Sensor
Networks (WSNs). The main target markets for WirelessHART communication
protocol are real-time process measurement and control applications. WirelessHART
is the first global wireless communication standard approved by IEC [4] which claims
to provide simple, reliable, and secure communication. A typical WirelessHART
based network is shown in Fig 2.2b. In addition to wireless communication capability,
WirelessHART also offers all the capabilities of the wired HART protocol.
Background
6
Figure 2.2b A typical WirelessHART network [1]
The architecture of HART protocol is shown in Fig 2.2c in comparison to the OSI
model. The architecture of the WirelessHART protocol can be seen in the right side of
Fig 2.2c, since it is a part of HART protocol.
The physical layer of WirelessHART protocol is based on the IEEE 802.15.4-2006
2.4GHz DSSS physical layer. The WirelessHART protocol operates in the 2400-
2483.5MHz license-free ISM band with a data rate of up to 250 Kbits/s. Two adjacent
channels in WirelessHART have 5MHz gap, and the channels are numbered from 11
to 26 [18].
The data link layer of WirelessHART is not compatible with IEEE 802.15.4-2006
DLL, rather it defines its own time-synchronized DLL. WirelessHART defines a strict
10ms time-slot and utilizes TDMA technology to provide collision free and
deterministic communications [17]. The concept of superframe is introduced to group
a sequence of consecutive time slots. Note a superframe is periodical, with the total
length of the member slots as the period. All superframes in a WirelessHART
network start from the ASN (absolution slot number) 0, the time when the network is
Background
7
first created. Each superframe then repeats itself along the time based on its period. In
WirelessHART, a transaction in a time slot is described by a vector:
{frame id, index, type, src addr, dst addr, channel offset}
Figure 2.2c The architecture of HART communication protocol [1]
where frame id identifies the specific superframe; index is the index of the slot in the
superframe; type indicates the type of the slot (transmit/receive/idle); src addr and dst
addr are the addresses of the source device and destination device, respectively;
channel offset provides the logical channel to be used in the transaction. To fine-tune
the channel usage, WirelessHART introduces the idea of channel blacklisting.
Channels affected Figure 2. WirelessHART Data Link Layer Architecture by
consistent interferences could be put in the black list. In this way, the network
administrator can disable the use of those channels in the black list totally. To support
channel hopping, each device maintains an active channel table. Due to channel
blacklisting, the table may have less than 16 entries. For a given slot and channel
offset, the actual channel is determined from the formula:
ActualChannel = (ChannelOffset + ASN) % NumChannels
The actual channel number is used as an index into the active channel table to get the
physical channel number. Since the ASN is increasing constantly, the same channel
offset may be mapped to different physical channels in different slots. This shows that
WirelessHART offers channel diversity and enhanced communication reliability.
Along with strict 10 ms timeslot, network-wide time synchronization, channel
hopping, channel blacklisting, WirelessHART DLL also offers industry-standard
AES-128 ciphers and keys.
The architecture of WirelessHART DLL is shown in Fig 2.2d. The WirelessHART
DLL has six major modules i.e.
Background
8
Interfaces
The interfaces between layers are meant for providing service primitives to the
corresponding higher layers e.g. the interface between the MAC and physical layer
defines the service primitives provided by the physical layer.
Timer
Timer is a fundamental module in the WirelessHART standard since WirelessHART
offers centralized network control/management. Its main purpose is to keep those
10ms timeslots in synchronization.
Communication Tables
Each network device maintains a collection of tables in the data link layer. The
information regarding the communication configurations and scheduling mechanism
is kept in the superframe table and link table. The neighbor table is meant for keeping
the list of neighbor nodes that the device can reach directly and the graph table is used
to collaborate with the network layer and record routing information.
Figure 2.2d WirelessHART data link layer architecture [3]
Link Scheduler
The link scheduler is meant for determining the next timeslot to be serviced based on
the schedule mechanism that is kept in the superframe table and link table. The
scheduler is complicated by such factors as transaction priorities, the link changes,
and the enabling and disabling of superframes. Every event that can affect link
scheduling will cause the link schedule to be re-assessed.
Message Handling Module
Background
9
The functionality of the message handling module is to buffer packets from the
network layer and physical layer separately.
State Machine
The state machine in the data link layer consists of three primary components: the
TDMA state machine, the XMIT and RECV engines. The TDMA state machine is
meant for executing the transaction in a slot and synchronizing the timer clock. The
XMIT and RECV engines deal with the hardware directly, which send and receive
packets over the transceiver [3].
The network layer and transport layer provide secure and reliable end-to-end
communication for network devices. The overall design of the network layer and
transport layer is depicted in Fig 2.2e.
Figure 2.2e WirelessHART Network Layer [3]
The basic elements of a typical WirelessHART network include: Field Devices,
Handhelds, a gateway and a network manager. WirelessHART is centrally controlled
by the network manager. Network manager is responsible for configuring the
network, scheduling and managing communication between WirelessHART devices.
To support the mesh communication technology, Each WirelessHART device is
required to be capable of forwarding packets on behalf of other devices since it is a
requirement for mesh communication technology. There are three routing protocols
defined in the WirelessHART standard i.e. source routing, graph routing, and
superframe routing [3].
The application layer is the highest level layer in WirelessHART standard that
provides infrastructure for building applications for WirelessHART based networks.
In the WirelessHART standard, the communication between the devices and gateway
is based on commands and responses. The responsibilities of the application layer
Background
10
include parsing the message content, extracting the command number, executing the
specified command, and generating responses [3].
2.3 Scheduling in WirelessHART
Since the data link layer (DLL) in WirelessHART is time-synchronized.
WirelessHART defines a strict 10ms time slot and utilizes TDMA technology to
provide collision free and deterministic communications. Transactions happen in
these 10ms time slots which are further divided as depicted in Fig 2.3a, each division
of the 10ms timeslot is used for a specific purpose. Infrastructure for time-
synchronization is defined in the WirelessHART standard but it is not defined that
which device should do the transaction in which time slot through which
communication channel? Thus a schedule (or scheduling algorithm) is required to
assign time slots of a communication channel for particular transactions for particular
devices.
Figure 2.3a WirelessHART slot timing
2.3.1 Location of Scheduling Algorithm in WirelessHART
The main scheduling algorithm resides in the network manager by using which the
network manager defines schedules. These schedules are broadcasted to the field
devices by using control and configuration messages. Field devices maintain these
schedules using link tables and superframe tables.
2.3.2 Requirements for Scheduling
The construction of the communication schedule is subject to several practical
constraints in WirelessHART networks [1]:
The maximum number of concurrent active channels is 16.
Each device can only be scheduled to TX/RX once in a slot.
Multiple devices can compete to transmit to the same device simultaneously
(in shared timeslot).
On a multi-hop path, early hops must be scheduled first.
The practical sample rates are defined as 2n sec (−2 ≤ n ≤ 9) from 250ms
(2−2
sec) to 8min and 32sec (29sec).
Background
11
2.3.3 Performance Measurers for MAC Scheduling Algorithms
Different performance measures are defined in literature for evaluating MAC
scheduling in WirelessHART mesh-networks. Some of these are:
Schedulable Ratio [2]
Buffer Size [2]
Scheduling Success Ratio [1]
Network Utilization [1]
Power Consumption
Packet Loss Rate or Delivery Rate
Throughput
Average and Maximum Delay
Power Consumption
2.4 State-Of-The-Art Scheduling Mechanisms
There are various scheduling mechanisms developed for WirelessHART network that
are given below, there might have also been developed other scheduling mechanisms
for the same purpose but I could have found these up till August 2011.
Optimal Branch-and-Bound (B&B) Scheduling [2]
Conflict-aware Least Laxity First (C-LLF) Scheduling [2]
Communication Schedules defined in [1]
Multihop multi-channel scheduling for wireless control in WirelessHART
networks [5]
Deadline-constrained transmission scheduling and data evacuation in
WirelessHART networks [6]
Optimal link scheduling and channel assignment for convergecast in linear
WirelessHART networks [7]
In this report the first three scheduling mechanisms/algorithms will be described since
these are the latest one and seem to be interesting but the details for any of the above
mentioned scheduling mechanism can be seen in the given references.
2.4.1 Optimal Branch-and-Bound (B&B) Scheduling
The B&B scheduling algorithm exploits the necessary condition established in
Theorem 2 (which states that for a set of flows F, let be the set of unscheduled
transmissions at slot s. If these transmissions are schedulable, then
[2]) to effectively discard infeasible branches in the
search space. The B&B scheduling algorithm guarantees to find a schedule whenever
a feasible one exits that makes it optimal and complete. The algorithm made decision
at every node by estimating an upper bound of the laxity of the schedule that the node
Background
12
may lead to. The laxity of a packet is “its remaining time slots minus its remaining
number of transmissions”, and the laxity of a schedule is “the minimum laxity among
all packets”. According to Theorem 2, for transmissions to be scheduled on or after
slot s, following is an upper bound (UB) of its schedule’s laxity:
(2.4a)
A global lower bound (LB) of schedule laxity 0 is maintained by the search. Two
decisions are made i.e. unschedulable or may be schedulable by computing UB at a
node (using Equation 2.4a). If UB < LB at a node, it is guaranteed that this node will
not lead to any feasible schedule and, hence, is discarded without further
consideration. And if UB LB, then this node may lead to a feasible solution and,
hence, is expanded further. The algorithm terminates as soon as it finds a feasible
complete schedule that meets all deadlines. If the original problem is infeasible, the
algorithm will also terminate as soon as it determines that this is the case [2].
2.4.2 Conflict-aware Least Laxity First (C-LLF) Scheduling
Conflict-aware least laxity first (C-LLF) scheduling is suitable for dynamic
environments where network topology changes frequently. In traditional Least Laxity
First (LLF) have been developed without taking, conflicts between transmissions, into
consideration and these are effective in end-to-end real-time scheduling over wired
networks, but in WSNs e.g. in WirelessHART based networks transmission conflicts
are highly expected that can easily decrease the effectiveness of the communication
schedule. Moreover, different nodes experience different degree of conflicts as
different nodes have different number of neighbors in a routing graph. The gateway
and the nodes with high connectivity in the routing graph tend to experience
significantly higher degrees of conflicts. Hence, scheduling algorithms for
WirelessHART networks must be cognizant of conflicts between transmissions [2].
Based on this key insight into the WirelessHART networks, an efficient scheduling
policy called Conflict-aware Least Laxity First (C-LLF) have been developed. It uses
conflict-aware laxity of every released transmission as the decision variable. The
conflict-aware laxity of a transmission is determined by considering the length of time
windows in which the transmission must be scheduled as well as the potential
conflicts that the transmission may experience in these windows. That is, the approach
combines LLF and the degree of conflicts associated with a transmission. Thus, it can
schedule a transmission while the remaining ones are likely to retain the necessary
condition established in Theorem 2 (which states that for a set of flows F, let be the
set of unscheduled transmissions at slot s. If these transmissions are schedulable, then
[2]). Specifically, the algorithm identifies some critical
time windows in which too many conflicting transmissions have to be scheduled,
thereby determining the criticality of each released transmission. Criticality of a
transmission is quantified by its conflict-aware laxity. Transmissions exhibiting lower
conflict-aware laxity are assessed to be more critical. C-LLF gives the highest priority
to the transmissions exhibiting lower conflict-aware laxity [2].
2.4.3 Communication Schedule Defined in [1]
This scheduling mechanism is based on the routing graphs that are defined in [1]. This
scheduling technique allows multiple devices to compete for the retry links to the
Background
13
same device, and split the traffic from one device among all its successors, thus
reduces the bandwidth allocation on each of them. By designing the communication
schedules on the successors so that their combination has the same communication
pattern as the original device, the global communication schedule is then spliced into
sub-schedules and distributed to the corresponding devices. These sub-schedules work
together and guarantee that the periodic process/control data between devices and the
Gateway can be forwarded through multi-hops in a timely manner.
The design philosophy for constructing this communication schedule is to spread out
the channel usage in the network as much as possible and to apply the Fastest Sample
Rate First policy1 (FSRF) to schedule the devices’ periodic publishing and control
data.
1Fastest Sample Rate First Policy (FSRF) ~ Rate Monotonic.
Design and Implementation
14
3 Design and Implementation
It is clear that optimized scheduling plays a significant role in enhancing the
reliability, network throughput, latency, and delay etc. of an IWSN but there is still
need of actual industry requirements analysis more precisely what is currently
required in the industry. Thus before designing the scheduling algorithm a detailed
requirement analysis has also been carried out by mainly focussing on process
automation industry.
Table 3a Typical Requirements for IWSNs in Process Automation Domain [9]
The process automation industry has mainly three types of data that need to be
communicated over an IWSN i.e. monitoring and supervision, closed loop control,
and interlocking and control. The wireless communication requirements for these data
types are tabulated in Table 3a.
3.1 Problem Formulation
The scheduling problem is formulated by defining the following properties for an
industrial wireless sensor network.
3.1.1 Payload
The payload of packets can contain four types of data i.e.
Monitoring and Supervision
For this type of data transmission, only sensors are involved and no actuators involved
means that only uplink communication is expected. Monitoring and supervision
sensors would be the 30% of the total number of sensor nodes. The update rate for
these sensors would be greater or equal to 1 sec. Allowed data loss is 1 packet/sec and
in case of data loss the allowed delay is 250ms or less. This type of data transmission
has the lowest priority.
Closed Loop Control
For this type of data transmission both uplink and downlink communication is
involved. These sensors and actuators would be the 70% of the total number of sensor
nodes out of which 52% would be the sensors and 18% would be the actuators. The
Design and Implementation
15
update rate for these sensors would be greater or equal to 500ms. Allowed data loss
delay is 250ms or less. This type of data transmission has the middle priority.
Interlocking
This type of data transmission would be event driven and would involve the closed
loop control actuators. Allowed data loss delay is 100ms or less. This type of data
transmission has the highest priority.
Network Management
In this data the information of network configuration and management is involved that
will be broadcasted time to time as required by the network manager.
The properties of the above mentioned data types are summarized in Table 3.1a.
Table 3.1a Properties of IWSN Data Traffic
Data Type % of total
Nodes
Sensors Actuators Update
Frequency
Data Loss Delay Priority
Monitoring &
Supervision
30 30% 0% 1 sec or greater
1 packet/sec 250 ms 4th
Closed Loop
Control
70 52% 18% 500 ms or
greater
very low 250 ms 3rd
Interlocking - - 18% or
less
event driven ideally no 100 ms 2nd
Network
Management
100 82% 18% event driven - - 1st
3.1.2 Network Topology
Routing is done according to the Cell-based Mesh Network topology. A routing table
would be available for building the schedule which contains routing details and link
cost for each link. Update rate of each node is also appended in the routing table for
the purpose of building the schedule. Several routing tables for different scenarios
have been generated on the basis of experimentation and mathematical models of
Cell-based Mesh Networks which are presented in [11]. Each cell will have one
backbone router and it should have a maximum of 30 sensors nodes in total.
3.1.3 Superframe
Superframe is defined as a group of consecutive time-slots with a fixed period i.e. the
total time taken by these time-slots. For instance in case of WirelessHART each time-
slot is 10 ms long, so if we consider a superframe of 25 time-slots then the period of
the superframe would be 250 ms.
Design and Implementation
16
3.1.4 Scheduling
Time-Slots
In this scheduling solution four types of 10ms duration time-slots in the superframe
are considered i.e.
Dedicated
These time-slots will be assigned for closed loop data transmission.
Shared
These time-slots will be assigned for monitoring & supervision data and
retransmissions of closed loop data.
Reserved
These time-slots will be used for broadcast data and network management.
Unused
These time-slots can be assigned for inter-cell communication or for any other
purpose.
3.2 Scheduling Algorithm
The scheduling is designed in such a way that it will always guaranty the periodic
closed loop data transmissions and event-driven interlocking data transmissions; this
is accomplished by defining reserved and shared time-slots with a period of 100ms.
For instance if a data packet couldn’t reach its destination within the dedicated time-
slot then the network manager can notify other nodes in the upcoming reserved slot of
the superframe about the retransmission and a shared slot or an unused slot will be
assigned for this retransmission; and in the event of interlocking first the network
manager will notify all the nodes using the reserved slots and then notify the actuators
in question using a reserved slot, a shared slot, or an unused slot within 100ms.
It is assumed that each cell can use only two channels at the maximum and there
would be four types of time-slots i.e. dedicated, shared, reserved, and unused. One
superframe will be assigned to each channel in the cell and the period of each
superframe would be the maximum update rate from the closed loop sensors. Thus
each cell would have two superframes working at the same time. Following is given a
scheduling algorithm that can be used to build a schedule for one cell; same procedure
can be repeated for all the cells in the network. The steps of the algorithm are;
1. Find the maximum update rate out of the closed loop control sensors; this
would be the period of the superframes.
2. Extract information of all the nodes related to one cell say Cell#01 from the
routing table.
Design and Implementation
17
3. Take the first generation of sensors in the cell; these sensors are directly
connected to the backbone router. This can be done using the routing table.
4. Initialize the first superframe with period that was found in step-1 by assigning
the first time-slot and one time-slot after every 90ms as reserved slot in the
superframe. The reason for having these time-slots with a period of 100ms is
to cope with the stringent requirement of the event-driven interlocking data
transmission which must be transmitted within 100ms. These time-slots can
also be used to notify for any retransmissions of closed loop data.
5. Assign two time slots after every reserved slot as shared slot in the
superframe. Normally these shared slots will be used for monitoring and
supervision data but in case of packet loss these can be assigned for
retransmissions of closed loop data as required by the network manager.
6. After doing steps 4 and 5 we have actually created 100ms time-cells within the
superframe, and each time-cell contains reserved, dedicated, shared, and may
also contain unused slots.
7. Assign dedicated time slots to the closed loop sensors according to rate
monotonic priority assignment (Rate monotonic is used here because it is
optimal [13] and static policy). Also assign dedicated time slots to the
monitoring sensors that are in the route of closed loop sensors.
8. While assigning time-slots to the routing nodes make sure that each of its
closed loop sensor child nodes is assigned one time-slot for transmission1 from
the routing node to the backbone router or to the parent node in case of second
superframe. In case of second superframe an additional time-slot to each
closed loop sensor child node will also be assigned to complete the
communication link up till first generation.
9. For more optimization, group the sensors and actuators that belong to closed
loop control with same update rate. Assign time slots to the sensors first then
to the actuators with the same update rate (here it is assumed that there is no
delay for the control loop processing and transmission between backbone
router (BR) and the network manager/gateway). In case of the control loop
processing delay and BR – Gateway transmission delay; the actuators will be
assigned at least two time slots after the sensors of the same update rate (in
this way we can give room of 20ms for processing and BR – Gateway
transmission).
Design and Implementation
18
RES
ERV
ED
SHA
RED
SHA
RED
RES
ERV
ED
SHA
RED
SHA
RED
RES
ERV
ED
SHA
RED
SHA
RED
RES
ERV
ED
SHA
RED
SHA
RED
FIRST SUPERFRAME
SECOND SUPERFRAME
Figure 3.2a Basic structure of superframes in a cell
10. The first superframe is done.
1The word transmission is used just for explaining purpose, the schedule is valid for both uplink as well as downlink communication.
11. Now take all the remaining generations of sensors in the cell, and initialize the
second superframe by assigning two shared slots as done in the first
superframe but the reserved slots will come just after the shared slots in this
superframe. The reserved slots are assigned after the shared slots to make sure
that a packet transmitted in a reserved slot of superframe can be transmitted to
higher generations by a routing node of first generation using the reserved slot
of second superframe within 100ms.
12. Repeat steps 6 – 9 to get the second superframe ready.
This scheduling mechanism is optimized for maximum 3 hops since in a Cell-based
Mesh network the maximum number of hops in a cell is 3. This algorithm can be used
to build schedule for more than 3 hops but the event-driven interlocking data
transmissions would not be guaranteed within 100ms in that case. The basic structure
of the two superframes in a cell is depicted in Fig 3.2a.
Design and Implementation
19
3.2.1 Flow Diagram
Design and Implementation
20
Design and Implementation
21
BR
M2
A4 M5
S3
S6 S7
Generation 1
Generation 2
Generation 3
M8
Figure 3.2b The hierarchy of nodes in a cell
To analyze the above mentioned scheduling algorithm let’s take the example of a cell
presented in Fig 3.2b. BR is the backbone router through which every node in the cell
communicates with the network manager. M is for monitoring nodes, S is for closed
loop sensor nodes, and A is for closed loop actuator nodes. In first superframe S3 will
be allocated a dedicated time-slot, M8 will not be assigned dedicated time-slot since
it’s a monitoring sensor but M2 will be assigned 3 time-slots since it has three closed
loop control child nodes i.e. A4, S6, and S7. In the second superframe A4 will be
assigned only one time-slot but M5 will be assigned four time-slots out of which two
meant for transmissions from child nodes to M5 and two meant for transmissions
from M5 to parent node i.e. M2.
3.3 Implementation
The implementation of the algorithm is done using LabVIEW [15] development
environment. The application program for generating the schedule is based on a GUI.
The GUI (top level VI1) reads the routing table from a .csv file, then generates the
schedule by using the given scheduling algorithm and outputs the schedule in the form
of two superframes. The GUI of the application is shown in Fig 3.3a. 1VI (Virtual Instrument) refers to a program that is written using LabVIEW programming language.
Design and Implementation
22
Figure 3.3a GUI for Calculating the Schedule
The VI hierarchy of the GUI based application program is shown in Fig 3.3b. The VI
hierarchy presents the dependencies of VIs among each other. The top level VI calls
three subVIs, the first subVI reads the routing table from a given .csv file. The other
two subVIs are meant for calculating the two superframes. The superframes are output
in the form of 2-D arrays. Each row of the array defines one time-slot and the columns
in the row have information about that particular time-slot. The first column is ‘slot
type’, the second is ‘source node’, and the third is ‘destination node’. Six types of
time-slots are defined within the scope of application program, i.e.
Unused = 0 (numeric representation)
Reserved = 1
Shared = 2
Dedicated = 3
TxFailure = 4 (this time-slot can be used to keep track of transmission
failures)
Routing = 5 (this time-slot is used by nodes for broadcast data)
Design and Implementation
23
Figure 3.3b VI hierarchy of the GUI based Application Program
The application program has been tested by testing and analyzing each of its sub-part
(SubVI). Superframes for two different cells were generated for testing purpose, and
then analyzed by comparing with the proposed scheduling algorithm and the routing
table and found no errors. Thus, this application program is reliable for doing the
performance analysis of the proposed algorithm.
Performance Evaluation
24
4 Performance Evaluation
The real-time scheduling problem for a WirelessHART network is NP-hard [2], so
heuristic approach will be used to analyze the reliability of the proposed scheduling
algorithm. Performance evaluation is done by considering two case studies from the
industry. For each case study, first the details about the cells of the scenario in
question are presented then the schedule is presented in tabular form. Afterwards the
throughput is calculated using the following equation;
Throughput =
The transmission failure is predicted using Poisson process that is given by the
following equation [14].
The value of is set to 0.25 for transmission failures.
4.1 Case-1: Oil Industry
In this case the area in question is in which there are 60 Sensor nodes in
total. The area is further divided into two cells and each cell is covered by one
backbone router i.e. BR1 and BR2 as shown in Fig 4.1a. Each cell will be utilizing
two frequency channels. Normally each backbone router should be responsible for
routing data of 30 sensor nodes.
Thus, most of the time each cell will have;
30% of 30 equals to 9 monitoring and supervision sensor nodes.
52% of 30 equals to 15 closed loop control sensor nodes.
18% of 30 equals to 6 closed loop actuator nodes.
The routing information for this scenario is given in Table 4.1a. The table header is
pretty straight forward except ‘A, S, or M’ column, this is meant for describing the
type of the sensor node i.e. A for closed loop actuators presented by numeric 1, S for
closed loop sensors presented by numeric 2, and M for monitoring sensors presented
by numeric 3 in the table.
Performance Evaluation
25
Figure 4.1a Distribution of sensor nodes in Case-1 [11]
Table 4.1a Routing table for Case-1
Node# Cell # Direct father
Generation# A, S, or M
Route cost
Update Period (ms)
9 1 36 2 2 2 500
10 1 67 2 3 2 1000
11 2 46 2 2 2 500
12 1 36 2 2 2 500
13 1 1 1 2 1 500
14 1 67 2 2 2 500
15 2 57 2 3 2 1000
16 1 1 1 2 1.02 500
17 2 60 2 2 2 500
18 2 2 1 2 1 500
19 2 43 2 2 2 500
20 2 28 2 1 2 500
21 1 48 2 2 2 500
22 2 2 1 3 1.12 1000
23 1 13 2 2 2 500
24 2 2 1 3 1 1000
25 1 64 3 2 3 500
26 2 28 2 3 2 1000
27 2 57 2 1 2 500
28 2 2 1 3 1 1000
29 1 48 2 2 2 500
30 2 2 1 3 1 1000
31 1 62 2 2 2 500
32 2 60 2 2 2 500
33 2 46 2 2 2 500
Performance Evaluation
26
34 1 67 2 1 2 500
35 2 50 2 2 2 500
36 1 1 1 2 1 500
37 1 36 2 3 2 1000
38 1 62 2 1 2 500
39 2 60 2 1 2 500
40 2 18 2 2 2 500
41 1 48 2 3 2 1000
42 2 2 1 1 1 500
43 2 2 1 2 1 500
44 1 1 1 2 1 500
45 1 1 1 3 1.29 1000
46 2 2 1 2 1 500
47 1 1 1 1 1 500
48 1 1 1 2 1 500
49 1 16 2 1 2.02 500
50 2 2 1 2 1 500
51 2 60 2 1 2 500
52 1 67 2 3 2 1000
53 2 22 2 3 2.12 1000
54 2 50 2 3 2 1000
55 1 62 2 1 2 500
56 1 13 2 3 2 1000
57 2 2 1 2 1 500
58 2 50 2 2 2 500
59 1 67 2 3 2 1000
60 2 2 1 2 1 500
61 1 16 2 3 2.02 1000
62 1 1 1 1 1 500
63 2 30 2 3 2 1000
64 1 36 2 2 2 500
65 2 28 2 2 2 500
66 1 36 2 2 2 500
67 1 1 1 3 1 1000
68 2 50 2 1 2 500
Performance Evaluation
27
4.1.1 Schedule for Cell-1
A detailed picture of the links in Cell-1 is shown in Fig 4.1b. After having all the
details about the Cell it’s time to build a schedule for it. The schedule is generated
using the GUI application program described in the previous chapter.
Figure 4.1b The spanning tree of Cell-1 of Case-1 [11]
The schedule is for Cell-1 is presented in Table 4.1b in the form of two superframes.
The first superframe will work in channel no. 1 and the second will work in channel
no. 2 of the cell.
Table 4.1b Schedule for Cell-1
Slot No.
SUPERFRAME 1 SUPERFRAME 2
Slot Type
Source Node Destination Node
Slot Type
Source Node Destination Node
1 1 0 0 3 9 36
2 2 0 0 2 0 0
3 2 0 0 2 0 0
4 3 44 1 1 0 0
5 3 13 1 3 12 36
6 3 13 1 3 14 67
7 3 16 1 3 21 48
8 3 16 1 3 23 13
9 3 36 1 3 29 48
10 3 36 1 3 31 62
11 1 0 0 3 34 67
12 2 0 0 2 0 0
13 2 0 0 2 0 0
14 3 47 1 1 0 0
Performance Evaluation
28
15 3 36 1 3 38 62
16 3 36 1 3 49 16
17 3 36 1 3 55 62
18 3 36 1 3 25 64
19 3 48 1 3 64 36
20 3 48 1 3 64 36
21 1 0 0 3 66 36
22 2 0 0 2 0 0
23 2 0 0 2 0 0
24 5 0 0 1 0 0
25 3 48 1 0 0 0
26 3 62 1 0 0 0
27 3 62 1 0 0 0
28 3 62 1 0 0 0
29 3 62 1 0 0 0
30 3 67 1 0 0 0
31 1 0 0 0 0 0
32 2 0 0 2 0 0
33 2 0 0 2 0 0
34 5 0 0 1 0 0
35 3 67 1 0 0 0
36 0 0 0 0 0 0
37 0 0 0 0 0 0
38 0 0 0 0 0 0
39 0 0 0 0 0 0
40 0 0 0 0 0 0
41 1 0 0 0 0 0
42 2 0 0 2 0 0
43 2 0 0 2 0 0
44 5 0 0 1 0 0
45 0 0 0 0 0 0
46 0 0 0 0 0 0
47 0 0 0 0 0 0
48 0 0 0 0 0 0
49 0 0 0 0 0 0
50 0 0 0 0 0 0
Performance Evaluation
29
The performance of the generated schedule can be analyzed from the plots presented
in Fig 4.1c and Fig 4.2d. In Fig 4.1c the successful packet transmissions are plotted
against required time-slots for these transmissions. In Fig 4.1d the Cell’s throughput is
plotted against required time-slots for the successful transmissions.
Figure 4.1c Successfully Transmitted Packets in Cell-1 of Case-1
Figure 4.1d Throughput in Cell-1 of Case-1
4.1.2 Schedule for Cell-2
A detailed picture of the links in Cell-2 is shown in Fig 4.1e. The schedule for Cell-2
is given in Table 4.1c and the results are shown in Fig 4.1f and Fig 4.1g.
Performance Evaluation
30
Figure 4.1e The spanning tree of Cell-2 in Case-1 [11]
Table 4.1c Schedule for Cell-2
Slot No.
SUPERFRAME 1 SUPERFRAME 2
Slot Type
Source Node Destination Node
Slot Type
Source Node Destination Node
1 1 0 0 3 11 46
2 2 0 0 2 0 0
3 2 0 0 2 0 0
4 3 42 2 1 0 0
5 3 18 2 3 17 60
6 3 18 2 3 19 43
7 3 28 2 3 27 57
8 3 28 2 3 32 60
9 3 43 2 3 20 28
10 3 43 2 3 33 46
11 1 0 0 3 35 50
12 2 0 0 2 0 0
13 2 0 0 2 0 0
14 5 0 0 1 0 0
15 3 46 2 3 39 60
16 3 46 2 3 40 18
17 3 46 2 3 51 60
18 3 50 2 3 65 28
Performance Evaluation
31
19 3 50 2 0 0 0
20 3 50 2 0 0 0
21 1 0 0 3 58 50
22 2 0 0 2 0 0
23 2 0 0 2 0 0
24 5 0 0 1 0 0
25 3 50 2 0 0 0
26 3 57 2 3 68 50
27 3 57 2 0 0 0
28 3 60 2 0 0 0
29 3 60 2 0 0 0
30 3 60 2 0 0 0
31 1 0 0 0 0 0
32 2 0 0 2 0 0
33 2 0 0 2 0 0
34 5 0 0 1 0 0
35 3 60 2 0 0 0
36 3 60 2 0 0 0
37 0 0 0 0 0 0
38 0 0 0 0 0 0
39 0 0 0 0 0 0
40 0 0 0 0 0 0
41 1 0 0 0 0 0
42 2 0 0 2 0 0
43 2 0 0 2 0 0
44 5 0 0 1 0 0
45 0 0 0 0 0 0
46 0 0 0 0 0 0
47 0 0 0 0 0 0
48 0 0 0 0 0 0
49 0 0 0 0 0 0
50 0 0 0 0 0 0
Performance Evaluation
32
Figure 4.1f Successfully Transmitted Packets in Cell-2 of Case-1
Figure 4.1g Throughput in Cell-2 of Case-1
4.2 Case-2: Paper and Pulp Industry
In this case the area in question is in which there are 100 Sensor nodes
in total. The area is further divided into three cells and each cell is covered by one
backbone router i.e. BR1, BR2, and BR3 as shown in Fig 4.2a. Each cell will be
utilizing two frequency channels. Normally each backbone router should be
responsible for routing data of 30 sensor nodes but in this case Cell-1, Cell-2 and
Cell-3 have 29, 42, and 29 nodes respectively. To focus on the results of the generated
schedule, the schedule tables for each Cell in this case are given in Appendix B. For
each cell first the distribution the nodes is depicted then result plots are shown.
Performance Evaluation
33
The routing information for this scenario is given in Table 4.2a.
Figure 4.2a Distribution of sensor nodes in Case-2 [11]
Table 4.2a Routing table for Case-2
Node# Cell # Direct father
Generation# A, S, or M
Route cost
Update Period (ms)
9 2 17 2 2 2 500
10 1 104 2 2 2 500
11 3 90 2 2 2 500
12 2 91 2 3 2 1000
13 1 1 1 3 1 1000
14 3 71 3 3 3 1000
15 2 82 2 1 2 500
16 2 108 2 2 2 500
17 2 2 1 1 1 500
18 2 92 2 1 2 500
19 1 1 1 1 1 500
20 3 86 3 2 3 500
21 2 108 2 3 2 1000
22 1 19 2 3 2 1000
23 3 93 2 3 2 1000
24 2 2 1 2 1 500
25 1 19 2 2 2 500
26 2 100 2 2 2 500
27 2 57 2 3 2 1000
28 1 1 1 1 1 500
29 3 66 2 2 2.19 500
30 3 93 2 1 2 500
31 1 28 2 2 2 500
Performance Evaluation
34
32 2 91 2 1 2 500
33 2 84 3 2 3 500
34 3 105 3 2 3 500
35 3 3 1 3 1 1000
36 1 1 1 2 1 500
37 1 52 2 3 2 1000
38 3 63 2 2 2 500
39 2 2 1 2 1 500
40 1 1 1 3 1 1000
41 2 108 2 3 2 1000
42 2 39 2 3 2 1000
43 1 19 2 1 2 500
44 1 36 2 3 2 1000
45 1 19 2 3 2 1000
46 1 19 2 2 2 500
47 1 13 2 2 2 500
48 2 91 2 2 2 500
49 1 104 2 1 2 500
50 1 62 2 2 2.07 500
51 2 103 3 2 3 500
52 1 1 1 3 1 1000
53 2 16 3 2 3 500
54 2 39 2 2 2 500
55 3 63 2 2 2 500
56 3 63 2 3 2.47 1000
57 2 2 1 3 1 1000
58 2 108 2 3 2 1000
59 3 93 2 2 2 500
60 2 2 1 2 1 500
61 2 91 2 2 2 500
62 1 1 1 2 1.07 500
63 3 3 1 2 1 500
64 1 94 2 3 2 1000
65 2 82 2 2 2 500
66 3 3 1 3 1.19 1000
67 1 13 2 2 2 500
68 3 78 2 2 2 500
69 1 52 2 1 2 500
70 3 90 2 3 2 1000
71 3 93 2 3 2 1000
72 2 82 2 2 2 500
73 1 104 2 2 2 500
74 1 13 2 3 2 1000
75 2 108 2 1 2 500
Performance Evaluation
35
76 1 28 2 2 2 500
77 2 88 3 2 3 500
78 3 3 1 2 1 500
79 3 55 3 2 3 500
80 2 92 2 3 2 1000
81 2 108 2 2 2 500
82 2 2 1 2 1 500
83 3 106 2 1 2 500
84 2 82 2 2 2 500
85 2 15 3 3 3 1000
86 3 35 2 1 2 500
87 3 78 2 2 2 500
88 2 24 2 2 2 500
89 1 1 1 2 1.03 500
90 3 3 1 3 1 1000
91 2 2 1 2 1 500
92 2 2 1 3 1 1000
93 3 3 1 3 1 1000
94 1 1 1 2 1 500
95 2 91 2 3 2 1000
96 2 80 3 3 3 1000
97 3 106 2 1 2 500
98 3 106 2 2 2 500
99 2 57 2 2 2 500
100 2 2 1 1 1 500
101 3 93 2 2 2 500
102 1 19 2 2 2 500
103 2 91 2 1 2 500
104 1 1 1 2 1 500
105 3 93 2 1 2 500
106 3 3 1 2 1 500
107 2 91 2 2 2 500
108 2 2 1 3 1 1000
4.2.1 Schedule for Cell-1
A detailed picture of the links in Cell-1 is shown in Fig 4.2b and the results are shown
in Fig 4.2c and Fig 4.2d.
Performance Evaluation
36
Figure 4.2b The spanning tree of Cell-1 in Case-2 [11]
Figure 4.2c Successfully Transmitted Packets in Cell-1 of Case-2
Performance Evaluation
37
Figure 4.2d Throughput in Cell-1 of Case-2
4.2.2 Schedule for Cell-2
A detailed picture of the links in Cell-2 is shown in Fig 4.2e and the results are shown
in Fig 4.2f and Fig 4.2g.
Figure 4.2e The spanning tree of Cell-2 in Case-2 [11]
Performance Evaluation
38
Figure 4.2f Successfully Transmitted Packets in Cell-2 of Case-2
Figure 4.2g Throughput in Cell-2 of Case-2
Performance Evaluation
39
4.2.3 Schedule for Cell-3
A detailed picture of the links in Cell-3 is shown in Fig 4.2h and the results are shown
in Fig 4.1i and Fig 4.1j.
Figure 4.2h The spanning tree of Cell-3 in Case-2 [11]
Figure 4.2i Successfully Transmitted Packets in Cell-3 of Case-2
Performance Evaluation
40
Figure 4.2j Throughput in Cell-3 of Case-2
Throughput is around 0.8 for different number of transmissions in almost every cell of
the presented case studies. Thus the proposed scheduling mechanism is suitable for
these industrial scenarios. Interestingly it also worked fine when the number of nodes
was 42 in Cell-2 of Case-2 and the successful packet transmissions crossed 400 by
utilizing 600 time-slots.
Conclusions and Future Work
41
5 Conclusions and Future Work
5.1 Conclusion
The results calculated in the previous chapter have proven that the proposed
mechanism is workable for different industrial scenarios by keeping the prominent
feature of transmitting interlocking data within 100ms. If we analyze the delays in the
different cases, we will reach the point that the delay is always less than or equal to
maximum update period of the closed loop control data i.e. the period of the
superframes. Latency is also less than or equal to the period of superframes.
It has been concluded that the proposed scheduling algorithm should be adopted for
the Cell-Based Mesh Networks, to have meet the stringent timing requirements of
IWSNs.
5.2 Future Work
Dynamic simulations of the proposed scheduling algorithm would be a beneficial
future work before experimenting it on a real-time test bed. Since both the Cell-Based
Mesh Network topology and the scheduling mechanism for it, are recently developed,
so it would be really handy if we develop a simulator that can do real-time simulations
and also analyze dynamic behavior of the network.
I had given it a try during my thesis work to develop a basis for such simulator using
National Instruments’ LabVIEW [15] development environment because using
LabVIEW real-time, we can develop real-time applications that can run on a real-time
target and the development time is considerably low as compared to other
programming languages. A GUI has been designed in an effort to start working on it
but due to shortage of time and more emphasis on developing a reliable scheduling
mechanism, it could not be completed. The designed GUI is shown in Fig 5.2a.
Figure 5.2a GUI Design for a Simulator
These are some of the efforts towards achieving the actual goal i.e. to develop
infrastructure for a reliable and secure Industrial Wireless Sensor Network.
References
42
6 References
[1] Song Han, Xiuming Zhu, Aloysius K. Mok, Deji Chen, Mark Nixon. “Reliable
and Real-time Communication in Industrial Wireless Mesh Networks”. The
University of Texas at Austin, Department of Computer Sciences. Report #TR-10-
34 (regular tech report). October 7th, 2010.
[2] Abusayeed Saifulah, Chenyang Lu, You Xu, and Yixin Chen, “Real-time
scheduling for WirelessHART networks”, in RTSS, 2010.
[3] Deji Chen, Mark Nixon, and Aloysius Mok. (2010) “WirelessHART™, Real-
Time Mesh Network for Industrial Automation”, Springer. ISBN 978-1-4419-
6046-7.
[4] “WirelessHART”,
http://www.hartcomm.org/protocol/wihart/wireless_technology.html (Acc. June
2011).
[5] Gabriella Fiore, Valeria Ercoli, Alf J. Isaksson, Krister Landernäs, and Maria
Domenica Di Benedetto, “Multihop multi-channel scheduling for wireless control
in WirelessHART networks”, in ETFA, 2009.
[6] Pablo Soldati, Haibo Zhang, and Mikael Johansson, “Deadline-constrained
transmission scheduling and data evacuation in wirelessHART networks”, in
Technical Report TRITA-EE 2008:060, 2008.
[7] Haibo Zhang, Pablo Soldati, and Mikael Johansson, “Optimal link scheduling and
channel assignment for convergecast in linear wirelessHART networks”, in
Technical Report TRITA-EE 2009:018, 2009.
[8] “Wireless Sensor Network”,
http://en.wikipedia.org/wiki/Wireless_sensor_network (Acc. November 2011).
[9] Johan Åkerberg, Mikael Gidlund, and Mats Björkman, “Future Research
Challenges in Wireless Sensor and Actuator Networks Targeting Industrial
Automation”, 2011.
[10] ISA100 http://www.isa.org/isa100 (Acc. Nov 2011).
[11] Yinchun Shen, “A Simulation Study of Cell-Based Mesh Network Topology and
Routing for Industrial Wireless Applications”, Master Thesis, Jönköping
University, December 2011.
[12] Youzhi Xu, Mikael Gidlund, Dong Yang, Wei Shen, and Tingting, “Robust
Routing in Cell-Based Mesh Networks”, Invention Disclosure Document, January
2011.
[13] Joel Goossens. “Scheduling of Hard Real-Time Periodic Systems with Various
Kinds of Deadline and Offset Constraints”. PhD Thesis, Universite Libre de
Bruxelles, 1999.
[14] Poisson Process, http://en.wikipedia.org/wiki/Poisson_process (Acc. November
2011).
[15] “LabVIEW Development Environment of National Instruments”,
http://www.ni.com/labview (Acc. November 2011).
[16] “Zigbee”, http://www.zigbee.org (Acc. November 2011).
References
43
[17] HART Communication Foundation, “TDMA Data Link Layer Specification”,
HCF_SPEC-075, Revision 1.1, May 17, 2008.
[18] HART Communication Foundation, “2.4GHz DSSS O-QPSK Physical Layer
Specification”, HCF_SPEC-065, Revision 1.0, September 01, 2007.
Appendices
44
7 Appendices
7.1 Appendix A State-of-the-Art Network Topologies
7.1.1 Star Topology
A typical star network topology is shown in Fig 7.1a. In a star network all the field
devices are connected directly to the gateway. It offers lower latency, simple
architecture and higher reliability.
M2
A4
M5
S3
S6
S7
M8
GATEWAY
Fig 7.1a Star Topology
7.1.2 Mesh Topology
In case of a mesh topology network all the nodes in the network must have the
capability to source and sink packets. The communication link is completed up till the
gateway by some of the nodes in the network that are directly connected to the
gateway. Mesh topology offers large area coverage but at the cost of increased no. of
hops which adds complexity to the routing and scheduling algorithms, lower
reliability, and higher latency.
7.1.3 Star-Mesh Topology
In case of star-mesh topology the network is divided into clusters and each cluster has
a head which is directly connected to the gateway. This topology is efficient than
mesh topology since it has lower no. of hops by the use of cluster heads, although the
failure of the cluster head will make the whole cluster fail. A typical star-mesh
network topology is shown in Fig 7.1b.
Appendices
45
Fig 7.1b Star-Mesh Topology
7.1.4 Hub-and-Spoke Topology
Hub-and-spoke topology is similar to star-mesh topology but in hub-and-spoke
topology only backbone routers can be cluster heads as compared to star-mesh
topology where any node can become a cluster head [12].
7.2 Appendix B Schedule Tables of Case-2
Table 7.2a Schedule for Cell-1
Slot No.
SUPERFRAME 1 SUPERFRAME 2
Slot Type
Source Node Destination Node
Slot Type
Source Node Destination Node
1 1 0 0 3 10 104
2 2 0 0 2 0 0
3 2 0 0 2 0 0
4 3 89 1 1 0 0
5 3 13 1 3 25 19
6 3 13 1 3 31 28
7 3 19 1 3 47 13
8 3 19 1 3 49 104
9 3 19 1 3 50 62
10 3 19 1 3 67 13
11 1 0 0 3 43 19
12 2 0 0 2 0 0
13 2 0 0 2 0 0
14 5 0 0 1 0 0
Appendices
46
15 3 19 1 3 69 52
16 3 28 1 3 46 19
17 3 28 1 3 73 104
18 3 28 1 3 102 19
19 3 36 1 3 76 28
20 3 52 1 0 0 0
21 1 0 0 0 0 0
22 2 0 0 2 0 0
23 2 0 0 2 0 0
24 5 0 0 1 0 0
25 3 62 1 0 0 0
26 3 62 1 0 0 0
27 3 94 1 0 0 0
28 3 104 1 0 0 0
29 3 104 1 0 0 0
30 3 104 1 0 0 0
31 1 0 0 0 0 0
32 2 0 0 2 0 0
33 2 0 0 2 0 0
34 5 0 0 1 0 0
35 3 104 1 0 0 0
36 0 0 0 0 0 0
37 0 0 0 0 0 0
38 0 0 0 0 0 0
39 0 0 0 0 0 0
40 0 0 0 0 0 0
41 1 0 0 0 0 0
42 2 0 0 2 0 0
43 2 0 0 2 0 0
44 5 0 0 1 0 0
45 0 0 0 0 0 0
46 0 0 0 0 0 0
47 0 0 0 0 0 0
48 0 0 0 0 0 0
49 0 0 0 0 0 0
50 0 0 0 0 0 0
Appendices
47
Table 7.2b Schedule for Cell-2
Slot No.
SUPERFRAME 1 SUPERFRAME 2
Slot Type
Source Node Destination Node
Slot Type
Source Node Destination Node
1 1 0 0 3 9 17
2 2 0 0 2 0 0
3 2 0 0 2 0 0
4 3 60 2 1 0 0
5 3 17 2 3 15 82
6 3 17 2 3 16 108
7 3 24 2 3 53 16
8 3 24 2 3 16 108
9 3 24 2 3 18 92
10 3 39 2 3 26 100
11 1 0 0 3 32 91
12 2 0 0 2 0 0
13 2 0 0 2 0 0
14 5 0 0 1 0 0
15 3 39 2 3 48 91
16 3 57 2 3 54 39
17 3 82 2 3 61 91
18 3 82 2 3 75 108
19 3 82 2 3 81 108
20 3 82 2 3 33 84
21 1 0 0 3 65 82
22 2 0 0 2 0 0
23 2 0 0 2 0 0
24 5 0 0 1 0 0
25 3 82 2 3 88 24
26 3 82 2 3 77 88
27 3 91 2 3 72 82
28 3 91 2 3 84 82
29 3 91 2 3 84 82
30 3 91 2 3 88 24
31 1 0 0 3 99 57
32 2 0 0 2 0 0
33 2 0 0 2 0 0
34 5 0 0 1 0 0
35 3 91 2 3 51 103
36 3 91 2 0 0 0
37 3 91 2 0 0 0
Appendices
48
38 3 92 2 3 103 91
39 3 100 2 3 103 91
40 3 100 2 3 107 91
41 1 0 0 0 0 0
42 2 0 0 2 0 0
43 2 0 0 2 0 0
44 5 0 0 1 0 0
45 3 108 2 0 0 0
46 3 108 2 0 0 0
47 3 108 2 0 0 0
48 3 108 2 0 0 0
49 0 0 0 0 0 0
50 0 0 0 0 0 0
Table 7.2c Schedule for Cell-3
Slot No.
SUPERFRAME 1 SUPERFRAME 2
Slot Type
Source Node Destination Node
Slot Type
Source Node Destination Node
1 1 0 0 3 11 90
2 2 0 0 2 0 0
3 2 0 0 2 0 0
4 5 0 0 1 0 0
5 3 35 3 3 29 66
6 3 35 3 3 30 93
7 3 63 3 3 79 55
8 3 63 3 3 59 93
9 3 63 3 3 68 78
10 3 63 3 3 83 106
11 1 0 0 3 38 63
12 2 0 0 2 0 0
13 2 0 0 2 0 0
14 5 0 0 1 0 0
15 3 66 3 3 55 63
16 3 78 3 3 55 63
17 3 78 3 3 86 35
18 3 78 3 3 20 86
19 3 90 3 3 86 35
20 3 93 3 3 87 78
21 1 0 0 3 97 106
22 2 0 0 2 0 0
23 2 0 0 2 0 0
24 5 0 0 1 0 0
Appendices
49
25 3 93 3 3 98 106
26 3 93 3 3 34 105
27 3 93 3 0 0 0
28 3 93 3 0 0 0
29 3 106 3 3 101 93
30 3 106 3 3 105 93
31 1 0 0 3 105 93
32 2 0 0 2 0 0
33 2 0 0 2 0 0
34 5 0 0 1 0 0
35 3 106 3 0 0 0
36 3 106 3 0 0 0
37 0 0 0 0 0 0
38 0 0 0 0 0 0
39 0 0 0 0 0 0
40 0 0 0 0 0 0
41 1 0 0 0 0 0
42 2 0 0 2 0 0
43 2 0 0 2 0 0
44 5 0 0 1 0 0
45 0 0 0 0 0 0
46 0 0 0 0 0 0
47 0 0 0 0 0 0
48 0 0 0 0 0 0
49 1 0 0 3 11 90
50 2 0 0 2 0 0