Proceedings of IC-BNMT2009

A TOKEN CYCLE SCHEDULING OF MAC

PROTOCOLS FOR TDMA BASED AIRBORNE AD
HOC NETWORK
Guo Zhonghai1, Chen Yongguang1, Duan Miyi2, Li Qiong3
1. Academy of Equipment Command & Technology Beijing 101416, China
2. Beijing Graphic Research Institute, Beijing 100029, China
3. Beijing Institute of Tracking and Telecommunication Technology, Beijing 100094, China
[email protected], [email protected]

(the gateway which connected the wireless ad hoc

Abstract network with the special process center). As for the
The MAC problem in a wireless network has satellite communication nodes, it may be used as
intrigued researchers for years. The MAC protocol long distant or emergency data communication. For
is also important for airborne ad hoc network in probably view of this airborne ad hoc network, it
order to reducing TDMA overhead, increasing may be depict in details in Fig1.
TDMA throughput, and decreasing End-to-End
delay. Our proposed Token Cycle Scheduling
described in this paper allow the exchange of
control information in an efficient and contention
free manner such that each participating node can,
in real time, converge to a consistent TDMA
schedule. The dynamic resource needs of
participating nodes can be rapidly accommodated
for QoS guarantees. The scheduling algorithms are
topology independent and require the nodes states
to assignment right slot to each node in order to
dynamically adapt to changing wireless network
topologies. Figure 1 Airborne ad hoc network

Keywords: MAC protocols; Token Cycle In the Fig 1, all the nodes can touch with each other
Scheduling; Airborne Ad Hoc Network as in there afforded communication distance.
Because of highly mobile in nature, all the nodes
1 Introduction can be category different clusters which have some
common clustering-related characteristic. From this
In recently years, the mobile ad hoc networks have point, it can be probably category into three-layered
been investigated intensely. Because of its auto- topology where the first layer is the satellite layer,
adaptive performance, great applications have been the second layer is the aircraft layer, and the third
adopted in many areas such as Vehicular Ad Hoc layer is the ground station layer. These layers could
networks (VANET), Emergency communication or effectively interact with each other using inter-layer
Reconnaissance in danger zone etc. links. Due to the high mobility of planes, the
network topology of airborne nodes falls into the
In this paper we present a candidate TDMA based Mobile Ad hoc Network (MANET) category. In
MAC protocols for highly mobile airborne network this network, traffic coordination is a constant
which can be self-forming, self-healing, adaptive challenge. Node topologies and transmission
data rate, Internet Protocol (IP)-compatible patterns are continuously changing, making
multiple-access network or in some special case wireless communications complicated and difficult
have the ability of jam resistant. There are several to maintain. Adaptive transmission scheduling must
kinds of available platforms or aircrafts for the be implemented in order to relieve the stresses
choice of airborne network nodes, including placed upon a MANET by packet collisions.
aeroplanes, helicopters, manned or unmanned Therefore the MAC protocols for all nodes to
aircrafts, airships etc. The satellite nodes or ground access the common channel resources will be the
station nodes can also be considered as an focus of this paper.
important part of the airborne network. The ground
station may be the control center or the relay station
___________________________________
978-1-4244-4591-2/09/$25.00 ©2009 IEEE

This paper is structured as follows: Section II 100Mbs. For the better of low latency, a half-
presents related work. Section III introduces the duplex slotted TDMA is adopted in the proposed
proposed Token Cycle Scheduling of TDMA based scheduling algorithm. A special channel is only
scheme. Section IV provides simulation of the used for control information and nodes state
proposed Token Cycle Scheduling. This is followed information which is used for assigning the rest
by a conclusion in Section V. wideband TDMA scheduling. For the whole,
because of avoiding collision among the nodes,
2 Related work high throughput and low latency can be achieved.
Every nodes in the network has there own Token
One approach where delay bounds can be slot only if it’s active. The Token slot is cycled in
“guaranteed” is time division multiple access all the active nodes. So the proposed scheme named
(TDMA), which has the ability to provide the Token Cycle Scheduling.
collision-free packet transmission regardless of the
traffic load. There have been many studies [1, 2] for 3.1 Assumptions Conditions
applying TDMA to distributed multi-hop self-
organizing network. However, most of them do not Each node in the network owns the directional and
take into consideration autonomous behaviors of omni-directional antenna, especially for the data
mobile nodes, and thus cannot assign time slots for transmission band to enhance spatial reuse. So the
new coming nodes. proposed scheduling also based on each node’s
position which can be calculated or acquired by
Kanzaki [3] has proposed a TDMA slot assignment GPS or control information. As for the time
protocol that assigns one of the unassigned slots to synchronization, this can be achieved by GPS or
a new coming node, but cannot assign more than each node’s broadcast information. As for the
one slot to a node whose message streams have rate whole network, a Cluster head which can connect
requirements and distance constraints. more nodes as far as its can is elected to manage the
network.
Dong [4] has proposed a real-time message model
to schedule message streams which need more than As for each node, they may be in only three states:
one slot. But it shows poor channel utilization and 1) Active: for some airborne nodes keep in active
acceptance ratio, and high time complexity because state because they are the main transmission
it must provide enough time slots for new coming backbone.
nodes and this causes a large number of unassigned 2) Non-active: maybe some ground stations or
slots which cannot satisfied the need of the other airborne nodes do not have the data to transmit. At
message streams either. this state, they just receive other node’s broadcast
information, in order to update the network state
In order to overcome the above deficiencies, a information, and to respond in the first time if other
number of topology-transparent (code-based) nodes call it.
scheduling methods have been proposed [7]. The 3) Power-off.
basic idea of the currently proposed topology-
transparent scheduling algorithms is for a node to 3.2 Packet Format
transmit in a number of time slots in each frame.
In the proposed Token Cycle Scheduling, a four
The limitations of the topology-transparent
time RTS and CTS handshaking mechanisms is
scheduling approaches described to date are that:
adopted. So there are several kinds of packet format,
The sender is unable to know which neighbor(s)
which included RTS packet, CTS packet, broadcast
can correctly receive the packet it sends in a
packet, ACK packet etc. Figure 2 shows some
particular slot; therefore, these methods cannot be
packet format in our protocol.
paired with RTS/CTS handshaking mechanisms
used to coordinate collision-free link activation
between a particular sender and a particular
receiver, or any process that requires the presence
of only one active transmitter in a two-hop
neighborhood.

3 Token cycle scheduling

The airborne ad hoc network is mainly used as the
communication backbone. It has wideband ability
to transmit high capacity and real time video or
picture data using Ku- or Ka-band BSM in the face
of constant topology changes and link quality Figure 2 Packet Format
fluctuation. Its transmission rate may reach up to


 CDMA code can enhance the ability of ant-
In the “FC” segment, it mainly contains general jamming.
information such as protocols edition etc.
The “Source. add” and “Dest. add” segment 2) If it has data request to transmit during its chief
always gives the source nodes and destination slot for each node, the RTS frame is used.
nodes.
As for the “Request information” segment, it 3) Each node in the network obtains other node
mainly contains the kind of QoS, the slot duration, state information from the broadcast information
and other transmission parameters. including theirs 1 or 2 hop distant Neighbor
The “Conform information” segment mainly Information List (NI-List) which contain node’s
used to inform the request node starting slot, the position, velocity, current using slot, etc.
directional antenna receive parameter, and the
destination node state including position, velocity, 4) As for the nodes in the cycle scheduling, there
usable timeslot etc. should be a cluster to manage the network, which is
The “Timeslot duration” segment contain the called cluster header. The fist node is elected as the
current node broadcasting duration of slot time to cluster header automatic at the beginning, and then
inform the whole network especial the neighbor is decided by which node could connect as more
nodes what the next broadcasting start timeslot, in node as it can.
order to avoid timeslot confliction.
The “Current state” mainly contains current node 5) In the time axis there should be a guard slot
state parameters in order to update state information between any two slots such as node D and E.
for its neighbor nodes. According to [8], using guard time can keep Space-
The “Slot information” notes what current node TDMA functioning without accurate
has recorded for other nodes using timeslot, such as synchronization. The maximum acceptable timing
its 1 or 2 hop distant neighbor nodes using timeslot difference between two adjacent nodes is Tguard, a
in the time axis. This is very useful for the new guard slot.
entry node to just the best joining in timeslot.
A) Operational Scheduling
The “FCS” segment use to check the frame.
Each node has it chief slot in turn or so called
3.3 Token Cycle Scheduling algorithm Token Cycle. During its “Token Cycle” slot, the
node can request to destination node and wait its
Several nodes have depicted in the Fig3, it is used response at CTS time slot. If the receives the CTS,
to explain how to work of the proposed scheme. then the two nodes can transmit data during theirs
handshaking time slot at the special data
transmission wideband. During the “Token Cycle”
B
A slot, other nodes can not transmit any signal and
just listen to the control channel in common CDMA
code to obtain other nodes state and the slot using
state. According to other node state and some
C D F
algorithm, each node can do corresponding reaction.
E New entry

When the next node listen to the Over sign, it

obtains the “Token” and begin to count. After a
guard time, if the node has data to transmit, it enters
ĂĂ E D B ĂĂ
the data request model; else, it just broadcast
Chief slot information and then transfers the “Token” to next
node.
Broadcast
Syn inf RTS
inf
CTS Reserved Over B) Joining the network
Figure 3 Example of proposed scheme When a node, e.g. F in Figure 3 wants to join the
network, firstly it should listens to the network
1) At the beginning of each node chief slot, there is broadcast information, and then gets the network
synchronization information contained for other operational parameters. Secondly choose the right
nodes to check out synchronization-time. Nodes neighborhood node (such as D) and the joining time
must synchronize each other using a slot which can not to interfere with other node
synchronization protocol [6]. For each node, a common communication and send the RTS frame
common CDMA code is used for synchronization to node D. After that node D will send the CTS
information and broadcast information frame and frame to node F and adjust some parameters and the
special CDMA code which every node has it time slot. When the next “Token” slot comes, the
special CDMA code for the RTS and CTS frame. node D will broadcast the new node’s information


and the changed time slot in order that node B can 4.2 Simulation Result
adjust time slot in time.
1˅Throughput Performance
C) Deviating from the network
Figure 4 presents the throughput performance of
When a node wants to deviate from the network, it proposed scheme in two states. It shows that as the
has two choices: 1) Waiting for the next chief slot mobility increased, the data packets throughput
and then broadcasting to the whole network special drop quickly especially with high traffic load. As
for its former node and latter node; 2) Doing for different kinds of QoS flow, it can also decrease
nothing at next chief slot. In this case, it latter node the throughput on order to satisfy their QoS
may have to use counter to just its start chief slot. requirement.

4 Simulation result
To evaluate the proposed Token Cycle Scheduling
performance, we take delay time, throughput and
network overhead as important three performance
metrics. In order to better know about our proposed
scheme, we compare the three performance metrics
with the IEEE 802.11e EDCA. The mobile nodes
may move at high or low speed, so we simulate
these two states at different arrive rate. All these
simulations are done at the OPNET simulator.

4.1 Simulation Setup

In our proposed airborne ad hoc network, there may
be several airborne platforms and ground station, so Figure 4 Throughput performance
as in the simulation, we supposed that there are 2) Delay Performance
about fifteen nodes in the air and five nodes on the
ground. The nodes are randomly distributed over an Figure 5 shows the delay performance. As the
area that is 500 km by 500 km. We used two nodes move faster, the delay time between two
different test cases: In the first case, each node's nodes may become longer. For one thing the
speed is randomly selected to be between 30 m/s to quickly movement make the listening to the control
50 m/s for each leg of the random-waypoint information more difficulty; for another the
movement; in the second case the nodes always position changed node may take interfere with other
move at 400 m/s. Different types of traffic are nodes. When the packet generation rate low, there
considered for our simulation, namely, voice, video, may be litter active nodes, so it may take more time
and data. When we consider different types of QoS to set up a network. It is also take long time to
traffic including voice and video, only the packets satisfy the QoS requirement when there is more
which satisfy their corresponding delay bound, are waiting packet.
counted to the aggregate throughput calculation.
For simplicity, in the simulation, the topology and
traffic pattern are generated randomly. The
simulation parameters are described on table 1.

The EDCA parameter set used for each traffic type

are the default values defined in [9].

Table 1 Simulation parameters

Parameter Values
Data frame payload size 8184 bits
Data channel rate 100Mbps
Control channel rate 2 Mbps
Velocity rang 10 m/s ~ 400 m/s
Simulation area 500h500 km Figure 5 Delay Performance
Node number 20
3) Network Availability
Frame duration 30 ms
Time slot duration
100 us Network availability is defined as the percentage of
Guard time duration 1 ms time that the network is able to provide a path for
data transport from any node in the network to


every other node. It provides a network-wide view [5] C. Zhu and M. S. Corson, “A Five-Phase
of the protocol's performance, whereas the Reservation Protocol (FPRP) for Mobile Ad
throughput and delay metrics only provide the hoc Networks”, Wireless Networks, vol. 7
protocol’s performance for a single route. Figure 6 issue 4: pages 371-384, 2001.
shows that as mobility increased, the network [6] J. Elson, L Girod, and D. Estrin, “Fine-
availability decreased and the required overhead Grained Network Time Synchronization using
increased (overhead increased as topology update Reference Broadcasts,” OSDI 2002, vol.36, pp.
rate increased). 147-163, Dec. 2002.
[7] Ji-Her Ju, and Victor O. K. Li, “An Optimal
Topology-Transparent Scheduling Method in
Multi-hop Packet Radio Networks,” in
IEEE/ACM Transactions on Networking, Vol.
6, No. 3, June 1998
[8] A. Ebner, H. Rohling, M. Lott and W.
Halfmann, “Decentralized slot
synchronization in highly dynamic ad hoc
networks,” in WPMC 2002, vol.2, pp. 494-
498, 2002.
[9] IEEE 802.11e/D11.0, Draft Supplement to
Part 11: Wireless LAN Medium Access
Control (MAC) and Physical layer (PHY)
specifications: Medium Access Control (MAC)
Enhancementsfor Quality of Service (QoS),
Figure 6 Network Availability performance Oct. 2004.
[10] B. Vandegriend, "Finding Hamiltonian Cycles:
5 Conclutions Algorithms, Graphs and Performance," Master
thesis, Department of Computing Science,
This paper has proposed a novel MAC scheme in University of Alberta, Edmonton, Spring 1998.
airborne ad hoc network with TDMA based Token [11] J. Elson, L Girod, and D. Estrin, "Fine-
Cycle Scheduling. The simulation has shown that Grained Network Time Synchronization using
the proposed scheme is high-efficient MAC Reference Broadcasts," OSDI 2002, vol.36, pp.
protocols which can adapt to high mobile and 147-163, Dec. 2002.
broadband airborne network. The latter work [12] A. Ebner, H. Rohling, M. Lott and W.
should take more time to do the slot assignment Halfmann, "Decentralized slot
algorithmic to improve the performances. synchronization in highly dynamic ad hoc
networks," WPMC 2002, vol.2, pp. 494-498 ,
References 2002.
[13] W. Ye, J. Heidemann and D. Estrin,"An
[1] A. M. Lewis, and S. V. Pizzi, "Quality of Energy-Efficient MAC Protocol for Wireless
Service for Tactical Data Links: TDMA with Sensor Networks," INFOCOM 2002, pp.
Dynamic Scheduling," Proc. of 2005 Military 1567-1576, 2002.
Communications Conference (MILCOM
2005), IEEE Press, Oct. 2005, Vol. 4, pp.
2350-2359.
[2] K. Kang, W. Guo, and R. T. Liu, “A novel
transmission schedule algorithm for TDMA
wireless multihop networks,” Dianzi Keji
Daxue Xuebao/Journal of the University of
Electronic Science and Technology of China,
vol. 36, no. 2, pp. 173-175, 2007.
[3] A. Kanzaki, T. Hara, and S. Nishio, "An
Adaptive TDMA Slot Assignment Protocol in
Ad Hoc Sensor Networks," Proc. of the 2005
ACM Symposium on Applied Computing,
ACM Press, 2005, pp. 1160-1165.
[4] L. Dong, R. Melhem, and D. Mosse,
“Scheduling Algorithms for Dynamic
Message Streams with Distance Constraints in
TDMA Protocol,” Proc. of the 12th Euromicro
Conference on Real-Time Systems, 2000, pp.
239-246.