Ant-Based Topology Convergence Algorithms for Resource Management in VANETs.

Full text

Page 1

Ant-Based Topology Convergence Algorithms
for Resource Management in VANETs
Frank Chiang, Zenon Chaczko, Johnson Agbinya, and Robin Braun
Faculty of Engineering, University of Technology Sydney,
Broadway, NSW 2007, Australia
frankj@eng.uts.edu.au
Abstract. FrequentchangescausedbyIP-connectivityanduser-oriented
services in Inter-Vehicular Communication Networks (VCNs) set great
challengestoconstructreliable, secureandfast convergedtopology formed
by trusted mobile nodes and links. In this paper, based on a new metric
for network performance called topology convergence and a new Object-
Oriented Management Information Base - active MIB (O:MIB), we pro-
pose an ant-based topology convergence algorithm that applies the swarm
intelligencemetaphortofindthenear-optimalconvergedtopologyinVCNs
which maximizes system performance and guarantee a further sustainable
and maintainable system topology toachieveQuality ofService(QoS) and
system throughput.Thisalgorithm is essentially a distributedapproach in
that each node collects information from local neighbor nodes by invoking
themethodsfromeachlocalized O:MIB,throughthesendingandreceiving
of ant packets from each active node, to find the appropriate nodes to con-
struct a routing path. Simulation results show this approach can lead to a
fast converged topology with regards to multiple optimization objectives,
as well as scale to network sizes and service demands.
1Introduction
Vehicular Ad-hoc Communication Networks (VANETs) are becoming an active
area in telecommunication research community [1]. As a Wireless LAN incorpo-
rating Mobile Ad-hoc Networks (MANETs), VCNs target the design and deploy-
ment of the QoS-assured, secure and high-speed communication connectivities
on mobile platforms such as buses, trains, cars, ships, marine squad of battle-
fields, where dispersed portable devices are enabled to establish on-demand per-
vasive communications in decentralized manners. Inter-vehicular, Intra-vehicular
ad-hoc mobile networks and vehicular-to-Internet communication networks are
indispensable parts for the envisaged ubiquitous communication scenario, this
adds more alternative ways for internetworking of WLANs, Satellite and Cellu-
lar systems. Currently, vehicular communications are typically required to: (1)
construct and maintain QoS-assured links, treat dynamic/static workload fairly
and further assign reasonable resources to support user-oriented services; (2)
provide acceptable reliability; (3) secure communication links and provide fault-
tolerance and self-restoration mechanism; (4) scale to future expected network
expansions and growths.
R. Moreno-D´ ıaz et al. (Eds.): EUROCAST 2007, LNCS 4739, pp. 992–1000, 2007.
c ? Springer-Verlag Berlin Heidelberg 2007

Page 2

Ant-Based Topology Convergence Algorithms993
To incorporateall theseissues,topologyconvergenceaimstofind the converged
topology which maximizes system capacity and throughput by minimizing inter-
ference and management costs, optimizing node cooperation and efficiently utiliz-
ing battery power as much as possible to self-adapting the topology such that the
required QoS and security concerns are satisfied.
1.1Topology Control vs Topology Convergence
A huge body of literature has been dedicated to topology control. However,
topology convergence is different from the topology control. Topology control
targets the maintenance of the chosen topology in a wireless network by adjust-
ing transmission power at each node [2] and such that an objective function of the
transmission powers are optimized. While topology convergence targets the con-
vergence of mobile hotspots or Access points internetworking paths to an optimal
topology not only before but also while the maintenance phase is invoked. By
establishing/updating a fast, reliable, secure and cost-effective topology under
multiple QoS constraints and management costs, it must be able to autonomi-
cally survive any hazards, resist any malicious attack, show certain adaptability
when roaming from one wireless LAN to another type of wireless LAN. And
system throughput, data routing and security requirements are achieved after-
wards. Several successive stages leading to topology convergence include (1) dis-
tributed topology discovery (e.g., gateways discovery for mobile hotspots); (2)
multiple objectives optimization process; (3) behavior/pattern matching with a
known optimal topology; (4) configuration/reconfiguation/activation. However
the nature of the formation of a converged topology in complex, dynamic, het-
erogeneous environments is unknown yet. This paper seems to be a first attempt
in exploring this domain and carry out a preliminary simulation test to better
understand the contributions from node cooperations as well as to understand
the effects of the convergence on network performance.
1.2Research Question and Contributions
It is the authors’ belief that the formation of the converged topology should
relate to the (1) routing mechanism (2) Availability of IP-connectivity (3) Op-
timal on-demand service-instantiated topology (4) Security issues. Our main
research question is: how do we address these new Network Management chal-
lenges, factoring in the ingredients of a distributed environment, to construct
efficient, robust, secure, scalable and mobile VANETs that are capable of pro-
viding the ubiquitous connectivity required by multimedia applications such as
VoIP and TVoIP? We argue that autonomic management paradigm can be the
solution to scalability and QoS requirements for distributed VANETs.
Specifically, our goals are to address the unexplored nature of topology con-
vergence for VANETs by creating a new routing protocol - an ant-based topology
convergence algorithm based on resource orientated cost function, topped with
an efficient self-managed network management paradigm. This new routing pro-
tocol is unlike conventional shortest path-based routing algorithms such as DSR

Page 3

994F. Chiang et al.
[3]. The multimedia streaming experimental results test performance of the QoS
of routing protocols for VANETs. We believe that our 4/3 management structure
[4], combined with the notions of an proposed O:MIB and biomimetic learning
and adaptation strategies will help to resolve the topology convergence issue.
The remainder of the paper is organized as follows. Section 2 presents the ant
colony biological model. Section 3 describes our algorithm and its application
to learning and adaptation. As a validation test, the performance comparison
in Section 4 shows the effectiveness and robustness to guide QoS-assured topol-
ogy convergence process in VCNs with respects to the known testing metrics.
Conclusion and future development are stated in section 5.
2 Ant Colony Biological Model
2.1Ant Agent Behavior Model Mapping to Network OSI Model
The ant agent behavioral model is inspired from the foraging behavior of an
ant colony. The ant model is working between the networking layer and the
application layer in order to improve the routing performance of IP data packet in
VCNs. In doing so, this biologically-inspired scheme can be seamlessly integrated
into current Internet infrastructure economically and flexibly.
A fully devised mobile vehicle having wireless bi-directional connectivity to
Base Station/Satellite ca classified as a node. Each node in VCNs contains one
ant colony, which can be classified into 4 parts: (1) packets queue repository, (2)
application/service repository, (3) communication channels and (4) Heterarchy
ant agents in the container.
The structure of an ant nest is shown in Figure 1. The packet queues reposi-
tory is the interface to the PHY and MAC layer while application/service queue
repository is the interface to the application layer. All the packets are sent out
or received in through the ant nest interfaces. The information flows go through
the environment inside and outside ant colony all the time, which facilitates ant
agents to communicate with each other through communication channels. The
communication channels include two modes: (1) indirect stigmergic communi-
cation channel and (2) direct contact communications. The ant agent container
contains heterarchy ant colony, where ant colony is constructed via a heterarchi-
cal way but not in a hierarchical way as described by Wilson & Holldobler in
[5] and Dr´ eo in [6]. Heterarchical structure makes the ants communication with
each other close without the level restriction as of hierarchical network. Het-
erarchy functions as a bottom-up designed system, where the system is highly
self-organized without the centralized control, emergent properties are emerged
in that each entity operates by following simple rules, when the whole colony
population is self-organized to work together and finally exhibits desirable pat-
terns and behaviors to reach system goals. While hierarchical structure follows
a top-down design, communications are restricted to go through various levels

Page 4

Ant-Based Topology Convergence Algorithms995
Fig.1. Ant Nest Model
to reach the final destination, and most of the time, a centralized control may
be needed.
2.2Conventional MIB vs Proposed Active O:MIB
Inspired from Active Networks [7], we propose an active O:MIB which is different
than traditional MIB in terms of the structure and its functionality. Object-
Oriented paradigms are applied into the construction of conventional MIBs who
possess holonic hierachy1.
3Constrained Ant-Based TC Protocols
The off-line ACO optimization applied into service management domain has
been tested in our paper [10] . This paper attempts to try online optimization
instead in the Network layer. The information interchanging are done in real-time
mode by the localized autonomic entity - O:MIBs which reside on each network
devices and services. Each local O:MIB establishes an information environment
where an artificial ant colony lives. The network system consisting of many local
O:MIB comes into being an information ecosystem, which mimicks the behav-
iors and patterns of the nature ecosystem. Network activities such as IP packet
transmission, forwarding, management information collection, decision making
processes are all happening in this ecosystem. Recent work [11] shows that the
congestion in manhattan networks are often undesirably worse in that the large
number of vehicular access points aggregated into some areas such as the crossing
area. The performance of getting fast converged topology for communications
can be improved by the proposed ant-based algorithms, which works according
to the network status, traffic balancing behaviors can be expected.
1Due to the page limits, full details of O:MIB are available in our papers [8] [9].

Page 5

996 F. Chiang et al.
3.1
As we design in previous section, each vehicular node i could be either source
node or destination node and functions asynchronously and independently from
each other, the node can interact with other nodes via sending and receiving ant
packets which is similar to the mechanism used in active networks. This protocol
is driven by on-demand events which can be roughly classified into four stages:
(1) Initialized by user-oriented on-demand service requests (2) Checking local
active O:MIB. (2) Preconfigured periods of ant packets sending and receiving.
(3) Update pheromone trails (4) Start another round of searching cycle in case
of unexpected disruptive events. Each node i needs to maintain a pheromone ta-
ble which contains time-varying pheromone value τ(CEij,t) for different CEij(·)
level. We use three different levels ω1(CEij(·)), ω2(CEij(·)),ω3(CEij(·)) to sim-
ply distinguish them as the weights for pheromone calculations and set the rules
accordingly for pheromone updating process. Table 1 shows the pheromone trails
for node 1 with its all neighbor nodes.
Ant-Based TC Protocol
Table 1. Pheromone Table
Node n1 Neighbour n2
Cω1
Eij
Cω2
Eij
Cω3
Eij
Neighbour n3
Neighbour n4
Neighbour n5
Neighbour n6
τn1,n2(Cω1
τn1,n2(Cω2
τn1,n2(Cω3
E1,2,t) τn1,n3(Cω1
E1,2,t) τn1,n3(Cω2
E1,2,t) τn1,n3(Cω3
E1,3,t) τn1,n4(Cω1
E1,3,t) τn1,n4(Cω2
E1,3,t) τn1,n4(Cω3
E1,4,t) τn1,n5(Cω1
E1,4,t) τn1,n5(Cω2
E1,4,t) τn1,n5(Cω3
E1,5,t) τn1,n6(Cω1
E1,5,t) τn1,n6(Cω2
E1,5,t) τn1,n6(Cω3
E1,6,t)
E1,6,t)
E1,6,t)
The edge parameter CE(i,j)between vehicles denote a function of multiple
factors that the topology convergence process needs to cover. The optimal con-
vergence process is then transformed into an ant traversing through the graphical
network where a certain number of vehicle nodes are visited until the stopping
criteria are reached. Moreover, each node is assumed to be traversed once at
most.
Suppose the ant is randomly placed on node i, the probability of an ant k
choosing the next adjacent node j as denoted in [12] is:
⎧
⎪
where ηi,jis called visibility and represents the heuristic desirability; U is the set
of neighbor nodes of node i that has not transpassed, and U is the set of feasible
components which can be potentially selected to fulfill a successful configuration
process. In our present method, information from holonic MIBs contribute to
the main part of ηi,j, denoted as ηi,j?
method for the topological convergence process are summarized as follows:
Pk(i,j) =
⎪
⎩
⎨
[τi,j]α[ηi,j]
?
β
l∈U
[τi,l]α
?
ηi,l
?β
0
if j ∈ U and l ?= j
otherwise
(1)
1
CE(i,j). The main steps of our proposed
Step 1) Initialize cost values and ACO parameters, such as predefined it-
eration times t, initial pheromone values τi,j(t) and stagnation conditions, etc.

Page 6

Ant-Based Topology Convergence Algorithms997
(1) Create M artificial ant agents for one ant colony, where M is usually set
to be equal to the number of nodes; Each ant colony is used to instantiate one
topology. (2) Set initial iteration variable t0= 0 ; set predefined iteration times
t=200; and set the initial pheromone value in all trails equal to a constant initial
value τi,j(0) = C. C should not be zero because a non-zero finite constant could
delay the process of evaporation into zero pheromone density so as to further
avoid the earlier local optima. (3) Dispatch each ant in this ant colony from
one node. The ant retrieves heuristic information from Holonic MIB, e.g., avail-
able bandwidth information, delay information, capacity and cost information,
connectivity information.
Step 2) Build the Solution
M ants start to follow the transition rule to build solutions asynchronously.
Meanwhile, heuristic information, as a transition constraint, ensures ants should
only traverse the feasible nodes in our problem domain before making a decision
of a move. Whenever the period of predefined time is due or the receiving of the
ant packets which require certain actions to be taken.
The edge parameter CE(i,j)(t) is calculated and usually less than 1. Until
now, all the heuristic information and edge vector values are determinate. The
probability of selecting next relaying vehicle at one iteration step is recorded
into a table. When the destination nodes are reached by all M ants, we consider
it as the completion of one round of iterations. There are 3 possibly successful
ant traversing scenarios: 1) Ant traverses through possible nodes until it reaches
destination node i and sleep there. 2) Ant reaches the destination nodes within
the maximum limits of hops counts H. 3) Suppose some node goes faulty, the
ants will be awaked up on this node to find an alternative node from the same
class of nodes as the faulty nodes, or the next available ant will find an alternative
node.
Let Pr(n−th Ant) represent the probability of being selected by the n-th
ant; the n-th ant in node i will select the next node j to visit with probability
equation (1)
Not all ants are able to finish a complete traverse path from product to re-
sources. For ants which are able to finish traverse paths during this iteration, we
match the components with the nodes on paths and record the minimum sum
value for CE(i,j)(t)
Step 3) Update pheromone trails
Upon receiving the ant packet from node i, equation (2) is applied to increase
pheromone value along the path. The insertions and reductions will influence
the searching process for the next iteration. We adopt the conventional ACO
pheromone update rule as:
– Intensify pheromone: an absolute amount of pheromone ?τk
to the existing pheromone values on the paths completed by k-th ant
?
0
i,j> 0 is added
?τk
i,j=
Q
Zmin(k)
if (i,j) ∈ Tk
if / ∈ Tk
(2)

Page 7

998F. Chiang et al.
Table 2. Speed-Density in Syndey CBD
VD(Vehicles/Km) Speed(Km/h)
40
80
120
240
58
52
40
10
where Tkare the paths having been traversed by k-th ant. Q is a constant,
more research on Q value have been done in Dorigo [13].
– Evaporate pheromone: When the wireless link life time is due or no more
packets go through, the evapration occurs as stated in equation 3.
τi,j= (1 − ρ) × τi,j
∀i,j ∈ [1 : m](3)
where ρ ∈ [0, 1) is the pheromone evaporation rate. τi,jis the current phero-
mone density on the trail a(i,j).
– Mutate pheromone update process
Step 4) Start another round of iterations
Before the predefined searching iteration times are reached, another round of
iterations always starts again, immediately after the previous round of iterations
ends. Meanwhile, the predefined iteration times t is increased by 1: t = t+1. Go
to step 2).
4Simulation
4.1Simulation Platform and Configuration Parameters
Vehicle Density (VD) in a lane is defined as the number of vehicles in a single
lane (n/km), hence, the maximum VD in a single lane can be calculated as:
Max{V D}lane=L
The recent extension work by Treiber and Helbing [14], who developed the
highway car following model further into a community intelligent driver model
(IDM) is given as:dυj
1 −
We implement the community mobility model into a simulation scenario that
emulates the real traffic patterns in community road with various driving states.
Based on this simulator, we apply our proposed scalable ant-based TC proto-
cols into this dynamic road condition and compare the performance of topology
convergence with other existing routing protocols such as DSR, AODV.
Metrics used in evaluating the performance of topology convergence for pro-
posed ant-based TC protocols: 1.) Packet Delivery Ratio; 2.) Message Overheads;
3.) Delay; 4.) System throughput and capacity.
l
dt= amax
?
?
υj
υdesired
?λ
−
?ddesired(υj,?υj)
dj
?2?
.

Page 8

Ant-Based Topology Convergence Algorithms999
4.2 Experiment Results
As for the conventional DSR and AODV, we consider the 3 performance mea-
sures: (1) Packet Delivery Ratio, (2) Message Overheads, (3) end-to-end Delays.
They are shown in the following figures. We compare the proposed TC algo-
rithm based on those metrics and prove it can efficiently improve the packet
delivery ratio and system throughput/capacity, and robustness and scalability
can be achieved too. Figure 2 and Figure 3 presents the simulation comparison
results on Packet Delivery Ratio (PDR) and end-to-end average delay between
our proposed scheme and AODV routing scheme.
100 200 300400500 600 700800 900 1000 11001200
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85

Ant-Based TC algorithms
AODV
Fig.2. Comparison of PDR
100 200300400500 6007008009001000 1100 1200
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5

Ant-Based TC algorithms
AODV
Fig.3. Comparison of Average Packet
Delay
5Conclusion and Future Work
This paper proposed ant-based topology convergence algorithm to guide QoS-
assured vehicular communication in future complex wireless networks. An innov-
ative active O:MIB is designed and implemented. The efficiency of the approach
outperform the current DSR and AODV. Future work includes: (1)Bi-directional
vehicular communications to be implemented into our simulator; (2) Scalability
and security issues to be taken into consideration.
References
1. Chisalita, I., Shahmehri, N.: A novel architecture for supporting vehicular commu-
nication. In: Proceedings of IEEE 56th Vehicular Technology Conference, vol. 2,
pp. 1002–1006 (2002)
2. Ramanathan, R., Rosales-Hain, R.: Topology control of multihop wireless networks
using transmit power adjustment. In: Proceedings of IEEE Nineteenth Annual
Joint Conference of the IEEE Computer and Communications Societies, vol. 2, pp.
404–413 (2000)
3. Johnson, D.B., Maltz, D.A.: Dynamic source routing in ad hoc wireless networks.
Mobile Computing 5, 153–181 (1996)

Page 9

1000F. Chiang et al.
4. Chiang, F., Braun, R.: Towards a management paradigm with a constrained bench-
mark for autonomic communications. In: Cheung, Y.-m. (ed.) Proceedings of the
2006 International Conference on Computational Intelligence and Security, vol. I,
pp. 520–523 (2006)
5. Wilson, E.O., Holldobler, B.: Dense heterarchy and mass communication as the
basis of organization in ant colonies. Trend in Ecology and Evolution 3, 65–68
(1988)
6. Dreo, J., Siarry, P.: A new ant colony algorithm using the heterarchical concept
aimed at optimization of multiminima continuous functions. In: Dorigo, M., Di
Caro, G.A., Sampels, M. (eds.) Ant Algorithms. LNCS, vol. 2463, Springer, Hei-
delberg (2002)
7. Bush, S.F., Kulkarni, A.: Active Networks and Active Network Management: A
Proactive Management Framework. Kluwer Academic/Plenum Publishers (2001)
8. Chiang, F.: Risk and vulnerability assessment of secure autonomic communication
networks. In: Wireless 2007. Proceedings of the 2nd International Conference on
Wireless Broadband and Ultra Wideband Communication (2007)
9. Chaczko, Z., Chiang, F., Braun, R.: Active mib: Addressing challenges of wireless
mesh networks. Lecture Notes in Computer Sciences (2007)
10. Chiang, F., et al.: Self-configuration of network services with nature-inspired learn-
ing and adaptation. Journal of Network and Systems Management 15, 87–116
(2007)
11. Guzman, G.N., Agbinya, J.: Topology analysis of moving wireless networks. In:
Auswireless 2006 (2006)
12. Bonabeau, E., Dorigo, M., Theraulaz, G.: Swarm Intelligence: From Natural to
Artificial Systems. Santa Fe Institute Studies in the Sciences of Complexity. Oxford
University Press, Oxford (1999)
13. Dorigo, M., Stuzle, T.: Ant Colony Optimization. MIT Press, Cambridge (2004)
14. Treiber, M., Hennecke, A., Helbing, D.: Congested traffic states in empirical ob-
servations and microscopic simulations. Physical Review E 62 1805 (2000)