Novel IoT Applications Enabled by TCNet: Trellis Coded Network

Diogo F. Lima Filho

and José Roberto Amazonas

2,3

Universidade Paulista – UNIP, Rua Dr. Bacelar, 1212, Vila Clementino, São Paulo, Brazil

Escola Politécnica da Universidade de São Paulo, Av. Prof. Luciano Gualberto – Tr. 3, 158, São Paulo, Brazil

Technical University of Catalonia – UPC, Department of Computer Architecture, Barcelona, Spain

Keywords: Internet of Things (IoT), Wireless Sensor Networks, Finite State Machine, Convolutional Codes, Trellis

Coded Network, Routing Algorithm.

Abstract: This work presents new results in routing in Wireless Sensor Networks, an important Infrastructure for the

Internet of Things architecture, using the new concept of Trellis Coded Network - TCNet. The TCNet is

based on the concept of convolutional codes and trellis decoder, that allow routing of data collected by

randomly distributed micro sensors in ad hoc networks scenarios. This model uses Mealy Machines or low

complexity Finite State Machines network nodes (“XOR” gates and shift registers), eliminating the use of

any routing tables enabling the implementation of important IoT applications as Sensor Network

Virtualization and in scenarios where clusters of nodes allow covering large areas of interest where the

sensors are distributed. The application of TCNet algorithm concepts in cases as VSNs and clustering is

facilitated due to the flexibility of TCNet to implement route management, becoming a tool to be adopted

by Sensor Infrastructure Providers aiming to deploy, for example, QoS-aware end-to-end services.

1 INTRODUCTION

Routing remains a challenge in today’s networks. It

is recognized that the major contributing factors are

the routing tables growth, constraints in the routers

technology and the limitations of today’s Internet

addressing architecture. Routing tables are populated

in routers and indicate the best next hop(s) for each

reachable destination along a route.

Considering that Wireless Sensor Networks

(WSNs) are an important infrastructure for the

Internet of Things (IoT) architecture, the interest in

using sensor networks in the same universe as IP

networks, although the sensor nodes have limited

hardware resources, this work explores an

innovative approach based on the concept of a

“Trellis Coded Network”- (TCNet), where the

foundations were introduced in previous works: (i)

“Implementation of QoS-aware routing protocols in

WSNs using the TCNet” (Lima and Amazonas

2012), where the network nodes are associated to the

states of a low complexity Finite State Machine

(FSM) and the links of a route are coded as the

transition of states of a convolutional code. The

routing discovery corresponds to finding the best

path in the convolutional code’s trellis; (ii) “A

Trellis Coded Networks-based approach to solve the

hidden and exposed nodes problems in WSN” (Lima

and Amazonas 2014), where it is explained how

TCNet innovates the decision making process of the

node itself, without the need for signaling messages

such as “Route Request”, “Route Reply” or the

“Request to Send (RTS)” and “Clear to Send (CTS)

to solve the hidden node problem that is known to

degrade the throughput of ad hoc networks due to

collisions, and the exposed node problem that results

in poor performance by wasting transmission

opportunities.

1.1 Related Work

Although some classic protocols as the Border

Gateway Protocol (BGP) (Rekhter et al. 2006),

Routing Information Protocol (RIP) (Hendrick,

1988), Ad-hoc On Demand Vector Routing (AODV)

(Perkins et al. 2003) and Open Shortest Path First

OSPF (Moy, 1988) address the scalability of today's

Internet routing system to consider the large number

of nodes that may be present in Low-Power and

Lossy Networks (LLNs) and IoT applications (IETF

620

Filho, D. and Roberto Amazonas, J.

Novel IoT Applications Enabled by TCNet: Trellis Coded Network.

DOI: 10.5220/0006772406200626

In Proceedings of the 20th International Conference on Enterprise Information Systems (ICEIS 2018), pages 620-626

ISBN: 978-989-758-298-1

ROLL 2009), a Working Group formed by the

Internet Engineering Task Force (IETF) in charge of

standardization and specifying the IP protocol

recognizes that factors like the routing tables

growth, constraints in routers technology and the

limitations of today's Internet addressing architecture

have driven the efforts in researching new

paradigms. Attempts to adapt the routing protocols

of infra-structured networks to cases of ad-hoc

networks are often inconsistent to address issues as:

frequent changes in topologies, poor link quality,

restricted bandwidth, and constraints on energy

resources. On the other hand, the ad-hoc and sensor

networks tend to proliferate as the number of smart

objects-based services on the Internet increase.

After this brief Introduction, Section 2 describes

the concept of a “Trellis Coded Network”- (TCNet)

to define the routing datagrams generated by each of

the nodes network, Section 3 presents the

performance of TCNet in terms of latency and

energy efficiency, Section 4 describes the

application of TCNet to the Sensor Network

Virtualization (SNV) and clustering scenarios and

Section 5 summarizes the conclusions and future

works.

2 TRELLIS CODED NETWORKS

The concept of a “Trellis Coded Network”- (TCNet)

changes conventional routing paradigms to enable

the development of QoS-aware packet forwarding

protocols in WSNs that are used to determine the

routing datagrams generated by each of the

network’s nodes offering the following advantages

that are compatible with the limited resources of

WSNs:

 Elimination of routing tables;

 Reduced latency by eliminating the route

request (RREQ) and route reply (RReply)

signaling packets employed, for example,

in AODV;

 Implicit self-recovery mechanism in case

of failure.

The model is based on finite automata (Hopcroft

and Ulman 1955) or FSMs defined by a ''cross''

function (k

/ out

) where a sequence of input

symbols {k

} generates a sequence of output codes

{out

} as shown in Figure 1. The k

(t) is the input

symbol received at time t and generates the output

code out

(t). It is also assumed that at time t, a

transition occurs at the FSM from state i to state j. In

TCNet each state represents a network node and the

transition state indicates that the frame information

must be sent from node i to node j.

It is then possible to generate a specific route

along a set of nodes, defined by a desired

optimization criterion (latency, packet loss,

throughput, cost), by shifting an input sequence {k

}

in the FSM of the route’s first node, and informed in

the frame as a TCNet label.

Figure 1: Network node modeled by a state of a FSM.

A network node modeled as a state allows the

application of FSM principles in order to exploit the

analogy between networks and state diagrams to

define paths between nodes. This enables each node

to have full knowledge of the network by

implementing a paths generator machine (MM) of

low complexity (''XOR'' gates and shift registers) as

shown in Figure 2.

Figure 2: Example of a MM with the input sequence {k

}

generating an output sequence out

(t) = (c

The TCNet architecture employs the Viterbi

algorithm (Proakis and Salehi 2008) proposed in

1967 for decoding convolutional codes based on the

trellis diagram to decide a sequence of branches to

be followed. The Viterbi algorithm decodes a

received sequence by evaluating the distance

between the sequence of received source symbols

and the weight of the path in the trellis, and

identifies the best sequence of branches as the one

that provides the minimum distance. This is done by

associating each branch with a number called branch

metrics, and looking for the path whose metrics sum

is minimum. This can be accomplished by means of

evaluating the maximum-likelihood (Gratzer, 1978)

and to produce an estimate of the received sequence

Novel IoT Applications Enabled by TCNet: Trellis Coded Network

621

of symbols.

Figure 3 shows how node-10 recognizes the

origin of the emitted sequence to establish a survivor

branch (probable partial route). Using the concepts

of the Viterbi Algorithm, the node in question

analyses the adjacent branches and, using the

Hamming-distance between the emitted sequence

and the respective weights of the branches, decides

in favor of the branch with minimum Hamming-

distance (hard decision-operation), node-00 (Haykin

and Moher 2009).

Figure 3: Example of the hard decision taken by node-10

when the sink node-00 emits the sequence out

(t)=(c

, c

)

= (1,1).

Using the described procedure, Figure 4 shows

the route established by the TCNet label using {k

}

= {1 1 0 0} and the MM depicted in Figure 2. It can

be observed that every node in the network is visited

in the order {(10), (11), (01), (00)}, i.e., at the end,

the frame returns to the sink node with the

information collected from every other node.

Figure 4: (a) Decoding trellis; (b) Route established by the

input sequence {k

} = {1 1 0 0} and the MM depicted in

Figure 2.

3 TCNet SIMULATION

EVALUATION

Consider the WSN scenario illustrated in Figure 5,

where the sink node initiates a query through a set of

sensor nodes in a predetermined order and also has

the function of Access Point to IP infra-structured

networks. The sink node initializes a frame loading

the WSN header field with the information

generated by the MM generator (out

(t)=(c

, c

)) and

transfers the input sequence ({k

}) to the TCNet

label field as shown in the Figure 6.

Figure 5: Illustration of a WSN in which the sink node

queries a set of sensor nodes. On the right, the TCNet

frame is shown.

Figure 6: Initialization of the TCNet frame by the sink

node: the input sequence {k

} is loaded on the TCNet label

field and the output sequence out

(t)=(c

, c

) is loaded on

the WSN header field.

The simulation environment used in this work is

the OMNeT++ based on C ++ (Varga, 2011) and

object oriented. This simulator is widely accepted by

the research community for being open software, has

been applied to the modeling of network traffic, and

as reference for comparisons with other available

frameworks.

Tests were done with an 8-node network, where

the sink node sends a query with CBR traffic to

verify the reachability of the nodes. Figure 7 shows

the used node’s model, configured by a MM with

rate k / n = 1/2 and the respective trellis decoder.

Figure 7: (a) MM with k/n=1/2, resulting output words

(n1, n2); (b) Trellis diagram corresponding to the MM.

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

622

3.1 The TCNet Performance Analysis

The same conditions of parameters and number of

nodes of the network were considered in order to

perform measurements of latency and energy

consumed by the network, as shown in Table 1. The

simulations considered a static scenario and the

worst-case where the performances for the most

physically remote and the most critical nodes of the

route were measured.

Table 1: Definition of the simulation scenario.

TCNet

AODV

Topology: 8 nodes

randomly positioned in an

area (1000m x 1000m) and

the sink node initiates

communication with the

other nodes by queries in

the network

Topology: 8 nodes randomly

positioned in an area (1000m

x 1000m) and the host (0) as

source node initiates

communication with the

other nodes of the network

using routes obtained

through flooding

Protocol: owner

Protocol: MAC 802.11g

Traffic: CBR with 512-byte

packets

Traffic: CBR with 512-byte

packets

Simulation time: 1 query

Simulation time: The 1st

flooding in the network, with

the negotiations (RREQ and

RRep) and the ACK

confirmation by the host (0)

3.1.1 Latency Efficiency

Figure 8 shows the results of the latency of different

nodes of the TCNet network in comparison with a

similar route for the AODV case. The latency has

been evaluated for the TCNet queries of the most

critical nodes (4, 6, 7 and 1), and the same nodes for

AODV route establishment mechanism. It can be

observed:

 In TCNet the latency increases as the

nodes correspond to the last positions of

the sequence k

(t), resulting a longer

processing time of the MM during the

decision making process of the target node.

 In AODV there is an initial delay in

establishing the route and a temporary

stabilization of the latency.

 In TCNet the worst-case latency

corresponds to 50% of the AODV latency,

in the considered scenario of an 8-node

network.

Figure 8: Comparison of TCNet and AODV latency for an

8-node network.

3.1.2 Energy Efficiency

The energy consumed in a WSN is a fundamental

parameter due to the limitations of the sources

(batteries) that are most often non-replaceable. This

work considered the energy consumed by the nodes

in the case of the TCNet algorithm, taking into

account the power distribution in the following

situations: Transmission (tx), Reception (rx),

Processing (proc) and Guard band (gb). The

following power values were adopted according to

the IEEE 802.11b standard:

 Transmission Power (P

): 2 mW;

 Reception Power (P

): 1 mW;

 Processing Power (P

proc

): 1 mW;

 Guard band (P

): not considered

The energy consumed by the node is given by

(1), which corresponds to the contribution of the

node to the total consumption of the network.

ΣE

(n)

proc

(1)

Table 2 shows the energy values for an 8 nodes

route considered in the simulation of TCNet

network.

Table 2: Individual contribution to the energy ΣE

(n)

consumed by the nodes of the TCNet network.

ΣE

(n)

Energy (Joule)

ΣE

(0)

4. 10

-4

ΣE

(4)

5. 10

-4

ΣE

(2)

6. 10

-4

ΣE

(5)

7. 10

-4

ΣE

(6)

8. 10

-4

ΣE

(7)

9. 10

-4

ΣE

(3)

10. 10

-4

ΣE

(1)

11. 10

-4

Novel IoT Applications Enabled by TCNet: Trellis Coded Network

623

The evaluation of the consumed energy by the

AODV considered the total time (ΣT

Lat

) taken by the

negotiations using the signaling (RREQ, RREP and

ACK) to establish the route to the destination node,

taking into account that in each AODV event energy

consumption occurs in the transmission, reception

and processing, respectively given by: E

, E

and

proc

. Thus the energy consumption for the AODV

network is given by (2):

aodv

= ΣT

Lat

proc

) (2)

In the comparison of energy consumption,

TCNet x AODV, it was considered an 8-node

network and a route going through all nodes to reach

the most distant node, i.e., node 7. The results are

shown in Figures 9 (a) and (b).

(a)

(b)

Figure 9: (a) Energy distributed among TCNet nodes in

relation to the energy consumed by the AODV; (b)

Comparison between the total energy consumed by the

mechanisms of TCNet x AODV.

The total energy consumed by the TCNet

corresponds to 75% of the energy consumed by the

AODV.

4 APPLICATIONS OF THE TCNet

ALGORITHM

4.1 TCNet in Scenarios of Sensor

Networks Virtualization

The use of the TCNet algorithm concept in cases of

Sensor Network Virtualization (SNV) (Anderson e

al 2005), (Chowdhury, 2009) is made easier due to

the flexibility of the TCNet algorithm in route

management applications.

The TCNet concept becomes a tool that can be

adopted by distinct Sensor Infrastructure Providers

(SInPs) to establish simultaneous end-to-end

services over a same infrastructure as shown in

Figure 10.

Figure 10: WSN virtualization environment scenario

enabled by the TCNet concept where two distinct SInPs

establish two simultaneous end-to-end services over the

same infrastructure.

The scenario presented in Figures 11 (a), (b) and

manages a heterogeneous set of sensors, over only

one infrastructure administered by a (SInP) and

controlled by a Gateway Router sensor located in the

sink. Figure 11 (c) shows an example of TCNet

managing different sets of sensors using queries

generated by different sequences corresponding to

specific quality requirements for each set of sensors.

In this example:

 The humidity sensors are visited by the

sequence: k

{1 1 0 0 0 0 0 0};

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

624

 The temperature sensors are visited by the

sequence: k

{1 0 1 0 0 0 0 0};

 The light sensors are visited by the

sequence: k

{1 1 1 1 0 0 0 0}.

Figure 11: (a) FSM with rate k/n = ½; (b) Trellis diagram

related to the FSM; (c) SInP sensors routing sets and

sequences related to the respective routes of the sensor

sets.

4.2 TCNet in Nodes Clusters Scenarios

Scenarios with large areas of interest to be covered

by WSNs suggest the subdivision of these areas into

clusters (Murthy and Manoj 2008). The use of

TCNet in these cases increases the alternative of

connections due to the self-conﬁguration of the

trellis, making unnecessary the use of signaling

protocols, as would be the case in ad-hoc networks

with AODV. Figure 12 shows a scenario with two

clusters based on trellis α and β and their respective

FSMs, allowing the construction of different routes.

Even occurring overlap of neighboring coverage

areas, the routes are independent and allows the

expansion in order to serve large areas of wireless

coverage.

Fig. 12. Clusters α and β have different settings of FSM

allowing the construction of independent routes.

The sink nodes, α-00 and β-00, allow the

interconnection of the clusters managing the routes

after the decision making process has been

performed by their respective trellis. The scenario

shown in the example of Figure 13 demonstrates

data collected from the nodes (α-01 and β-01),

belonging to the different clusters, and being

transmitted by their respective sinks to be

aggregated to the IP traffic by sink  performing

its gateway function.

Figure 13: Clusters corresponding to the FSMs and

showing a concatenation of the routes between the two

clusters.

5 CONCLUSIONS

In this work we have made a review of the TCNet

concept that enables the implementation of packets

forwarding procedures in limited processing,

storage, communication and energy resources

networks, as WSNs, without using routing tables. In

addition a comparative performance evaluation

between the TCNet and the AODV was made

showing that the TCNet outperforms the AODV

both in terms of worst-case latency and total energy

Novel IoT Applications Enabled by TCNet: Trellis Coded Network

625

consumption. It has also been demonstrated the

potential of TCNet to be used to implement sensor

virtualization networks and the management of

sensors clusters. These applications are important in

the IoT domain and show TCNet as an enabling

technology to tackle scalability and to offer different

levels of QoS.

The TCNet concept is a powerful tool that can be

adopted to face very challenging problems. As

future work we will demonstrate how TCNet can be

used to implement robust networks with self-

recovery properties in the presence of failures.

ACKNOWLEDGEMENTS

This work has been partially supported by the

Spanish Ministry of Economy and Competitiveness

under contract TEC2017 – 90034 - C2 - 1 - R

(ALLIANCE project) that receives funding from

FEDER.

REFERENCES

Anderson,T., Peterson, L., Shenker, S. and Tuner, 2005.

“Overcoming the internet impasse through

virtualization. Computer”, v. 38, p. 34–41.

Chowdhury,K., Zaheer, F. and Boutada. 2009. An identity

management framework for network virtualization.

IEEE INFOCOM, p. 34–41.

Gratzer,G., 1978. “General Lattice Theory”, 1

ed.,

Academic Press, Inc.

Haykin, S and Moher, M, 2009. Communication Systems,

5th ed., John Wiley & Sons, Inc.

Hendrick,C. L, 1988. “Routing Information Protocol,”

RFC 1058, available for download from

http://www.rfc-editor.org/rfc/rfc1058.txt.

Hopcroft, J .E. and Ulman,J. D., 1955. “Introduction to

Automata Theory”, Languages and Computation, 1st

ed. Addison Wesley.

IETF ROLL Working Group, 2009. “IETF Roll Working

Group routing over low-power and lossy networks,”

homepage: http://www.ietf.org/dyn/wg/charter/roll-

charter.html.

Lima, D. F. and Amazonas, J. R, 2012. “TCNet: Trellis

Coded Network – Implementation of QoS – aware

routing protocols in WSNs ”, IEEE Latincom

106445, http:www.ieeelatincom.org/2012/.

Lima, D. F., Amazonas, J.R., “A Trellis Coded Networks-

based approach to solve the hidden and exposed nodes

problems in WSN,” SICOTEL2014 Seminário Taller

Latinoamericano de Instrumentación, Control y

Telecomunicaciones, April 2014,

http://www.gituq.edu.co/memorias%20sicotel2014.pdf

Moy, J., 1988. “ Open Shortest Path First,” RFC 2328,

available for download from http:www.rfc-

editor.org//rfc/rfc2328.txt.

Murthy, C. R., Manoj, B.S., 2008 “Ad Hoc Wireless

Networks – Architectures and Protocols”. 6

th. Ed.

[S.l]: Prentice Hall.

Perkins, C, Belding, E. and Das S., 2003. “Ad hoc on-

demand distance vector-(AODV) routing,” RFC 3561,

available for download from http:www.rfc-

editor.org/rfc/rfc3561.txt.

Proakis,J.G., Salehi, M., 2008. Digital Communications,

5th ed., Mc Graw Hill.

Rekhter, Y, Li T. and Hares S., 2006. “Border Gateway

Protocol”, RFC 3271, available for download from

http://www.rfceditor.org/rfc/rfc4271.txt.

Varga, A.: “OMNeT++, 2011. Discrete Event Simulation

System”,

http://www.omnetpp.org/doc/manual/usman.html.

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

626