Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Wireless In-Building Communication

Using an Array of Microdevices

J-M. Dilhac, D. Dragomirescu.


LAAS-CNRS
7 avenue du colonel Roche, 31077 Toulouse CEDEX 4, FRANCE

Abstract- In this paper a new concept is proposed for routing microdevice network consists of numerous, simple, low-cost
communications between two terminals through a low level and highly unreliable components and interconnections. As a
protocol utilizing a network of periodically deployed microdevices result, it possesses very limited processing and energy
acting as autonomous nodes, without any supervising entity. capabilities together with high failure rates. Considering the
Nodes are neither primary sources of information nor routers. above, there are strong similarities with computing and
They basically act as repeaters with very limited processing networking using electronic textiles [4].
power. On-demand routing does not require frames: the
Consequently, new networking procedures, innovative
communication protocol works at the symbol level. In a first step,
to perform route discovery, the minimum propagation time Medium Access Control (MAC) and unusual transmission
between terminals is measured. Then, the nodes situated along protocol, differing strongly from what is found in conventional
the corresponding shortest route are activated. Finally, computer network, are needed.
information is exchanged using a full-duplex scheme. This paper is organized as follows. Section II describes the
MEMS network while section III outlines our proposed MAC
I. INTRODUCTION and communication protocol. Section IV considers some
preliminary issues related to our protocol performances.
The emergence of microelectromechanical systems In the following, we mostly ignore hardware
(MEMS) incorporating communication components has implementation issues, even if we mention a few practical
created enriched opportunities to design innovative systems for aspects. Our work focuses on the MAC and transmission
information transmission. Cubic-millimeter MEMS are control aspects only.
considered for deploying a massively (i.e. hundreds to
thousands dust motes) distributed sensor network, as in the II. PHYSICAL DESCRIPTION
University of California, Berkeley 's Smart Dust project [1].
Advances in microelectronics provide increasing A. Network topology
functionalities in smaller sizes, considering sensing, signal
processing, communications and energy management. The medium of transmission considered in this paper (see
However, in that kind of application, MEMS rely on a Fig. 1) is a two-dimensional (2-D) interconnection pattern, i.e.
central station to send and receive data: either there is a direct a surface periodically paved with microdevices acting as
communication between an entire field of MEMS to a base autonomous (both with respect to energy and the absence of a
station, or alternatively a multihop network is formed to allow supervising entity) nodes, thus assembled into a meshed
sensor information to propagate to a central collecting point multihop topology. This pattern, absent faults, does not
[2]. In that last case, intermediate nodes assist distant nodes to change. This 2-D structure, possibly flexible (but all nodes are
reach the central station. In these applications, MEMS still spaced exactly d apart) may therefore be embedded in
incorporate a sensor, and the purpose of the network is to three-dimensional (3-D) space. Consequently, this network is
collect information about the environment and to propagate it different from a sensor network, like the one described in
back to the collecting point, with some in-network processing section I, where the nodes are randomly distributed.
such as aggregation or compression [3]. It is there basically a The network surface may be of any size and pattern, either
wireless sensor network (even if most of the time, sensor fully paved by microdevices or with missing nodes at given
nodes will not have originating traffic to send), which does not positions thus corresponding in practice to physical obstacles.
support one or several independent point-to-point flows, while Missing nodes may also correspond either to nodes in idle state
in our application, the microsystems act as repeaters. for energy purposes or to some broken nodes or
The purpose of our paper is therefore to present a different interconnections.
and innovative networking strategy using distributed Nodes may be considered as low-cost simple components,
microdevice nodes. The goal of this strategy is to allow two that is silicon-based microsystems or even nanosystems. As
terminals to communicate using microdevices as nodes of a already said, nodes and communication links may be failure
maintenance-free, fault-tolerant network together with a low- prone giving either permanent or transient (when failure is
level transmission protocol. The objective is to identify and related to power management) faults.
create a minimum-length path of transmission, therefore In the next decades, non-silicon technologies (nano,
avoiding delay-spread associated with multipath propagation, molecular, polymers…) may be of interest with the added
while node functionality and overhead are kept minimum. advantage of allowing the processing of filament-type batteries
Indeed, a microdevice network is not a classic data network mixed with the wire lines [5].
exhibiting boundless capacity and processing power. Such a

IEEE Communications Society 3539 0-7803-8533-0/04/$20.00 (c) 2004 IEEE


The traffic will propagate through the network over several permit communication. The consequence is a multipath-free
hops, with limited intermediate processing, to larger- communication at the microscopic level.
processing nodes called "terminals" (see Fig. 1). As explained The nodes are capable of receiving and processing at most
below, the nodes act basically as repeaters at the symbol level: four simultaneous communications, and able to distinguish
they ignore, when used by the terminals, a possible frame between the originating locations. They can simultaneously
structure. The reason is that, to avoid the drawbacks associated emit up to four different messages in the four directions.
with complex information processing algorithms, such as However, a node has no a priori knowledge of the network and
power consumption, processing time, increased device size, we will continue ignoring its position within this network. More
need to devise a scheme with which the task performed by a precisely, it initially does not know which of its neighbours is
node is as simple as possible. Moreover, the nodes are not to or are closer to a given terminal. Moreover nodes have neither
be reprogrammed after being deployed even if the network physical nor logical address.
topology is modified. Nodes are neither primary sources of information nor routers.
A node is silent until a broadcast transmission activates it.
They can be regarded as automata, each one being associated
with its own clock.
They are more or less acting as repeaters with very limited
processing power and very little buffering capability. In the
following, they will be referred to as microrepeaters.

C. Terminals
d
The terminals are the primary sources of information. The
x energy-unconstrained terminals considered in this paper may
d be temporarily plugged anywhere within the network.
However their positions must not change during a
Figure 1: example of a communication network formed by microdevices
(diamonds). The distance between closest neighbours is d. Two terminals
communication phase. Moreover, these positions must
(crosses) are shown (see text). The connectivity space of microdevice labelled correspond to a node the functionality of which is then
x is shown. replaced by that of the terminal. However, like the
microrepeaters, the terminals ignore both their own positions
The nodes could communicate using wireless means. within the network, and initially its location with respect to the
Radiofrequency and optical communications could be other terminal(s). Therefore, messages cannot be addressed by
considered. As previously stressed, it is not our purpose to location. Nevertheless, terminals may possess a logical
consider in details the issues associated with the practical address. In the following, we will consider only two terminals,
implementation. However, as will be explained in the next labelled A and B.
paragraph, a node is supposed to communicate only with its
closest neighbours. Collimated optical beams might therefore III. COMMUNICATION PROTOCOL
be the solution of interest. Moreover, nodes being periodically
arranged, the network primary application is probably
associated with an in-building environment, the microdevices A. Introduction
being incorporated into the fabric: floors, ceilings, walls...
Alternatively, the nodes could be connected by electrically We describe here the transmission control scheme we
isolated wires twisted with yarns used in woven fabrics, that is propose. It is very different from what is found in sensor
one of the interconnection methods used for electronic textiles networks [3] where each node is both source and router, and
[4]. It is worth mentioning that this alternative is much more where routing tables and frames (packets) are used.
power efficient than wireless communication found in sensor Nevertheless, the network nodes collaborate as a whole to
networks [6]. deliver observations leading to robustness to the loss of
individual nodes.
B. Node functionality Our low-level communication protocol works at the
symbol level. We assume that an alphabet consisting of eight
The nodes constitute the infrastructure allowing the symbols is available. In a preliminary work [7], we used an
terminals to communicate. The bidirectional, full duplex alphabet size which was proportional to the network extent:
connectivity between nodes is static: a node can only this was obviously a strongly limiting factor, no more
communicate with its nearest neighbours at distance d, thus (in
impeding the improved protocol version presented here.
the example of Fig. 1) at most four nodes if all are present. The
The establishment of a communication channel goes
connectivity space is therefore a circle, radius of which is
through three steps, the first two steps corresponding to on-
slightly larger than d (see Fig. 1). In other words, the nodes
demand route discovery:
have anisotropic emission/reception features and an unbroken
line-of-sight is required between a node and its neighbour to - calculation of the shortest path length through the
measurement of the minimum propagation time T,

IEEE Communications Society 3540 0-7803-8533-0/04/$20.00 (c) 2004 IEEE


- activation of the microrepeaters along this path, last symbol has followed the shortest path as it exhibits the
- information transfer by exchange of time series data. minimum delay. Then, a last broadcast symbol labelled M3 is
The communication ends up by a reset phase. emitted by A in response to the reception of the first received
However, before describing this procedure, more details M2 symbol. The protocol is summarized in Fig. 2 and 3 at the
need to be given about network synchronization. terminal and microrepeater levels respectively.
(terminal A) (terminal B)
B. Node to node communication and network synchronization
start start
When a symbol is transmitted, either from a terminal, or
from a microrepeater to the nearest neighbours, the
corresponding information propagation time is labelled τ. M1 no
More precisely, τ is the time interval between the reception of M1 ?
one (or the simultaneous reception of up to four) symbol(s) by
a microrepeater, and the reception of the (or the simulataneous
reception of up to four) reemitted symbol(s) by the (up to four) no yes
neighbours of the initial microrepeater.
M2 ?
However, while being referred to as microrepeaters for M2
simplicity, basic information processing is done by the nodes
as the (re) emitted symbols may be different from the received yes
ones. Assuming that symbols could be (re)emitted after some
processing time tp which may vary with the task complexity (to
M3 no
keep the propagation time τ constant) the microrepeater must
restrain from transmitting from time of reception t to time t + tp
M3 ?
+ TIMEGUARD with
yes
tp + TIMEGUARD + tT = τ (1)

Figure 2: partial representation of transmission protocol of Mx symbols for


with tT being the true physical propagation time of symbols terminals. Emissions are identical in all directions.
between two nodes at a distance d from each other. As d is a
stands for emission and for reception.
constant, tT is independent of the node positions.
TIMEGUARD is updated after the completion of the local
information processing. A similar process applies to the As can be seen from Fig. 3, a microrepeater (including
terminals to determine the time interval between emissions sender) will emit a given Mx symbol towars all neighbours,
(see Fig. 4 in section III.c). but only once. This has two effects: first the Mx symbols will
It is worth mentioning that the practical application of (1) "diffuse" isotropically (through multiple hops) from the
necessitates accurate time interval measurements at the node emitting terminal or repeater. In other words, it is broadcasted
and terminal levels but, in return, this will result in to all of the microrepeaters. Second, in networks exhibiting
simultaneous emissions from all the active complex topologies, these symbols will not travel permanently
microrepeaters/terminals and therefore, as seen below, in in the network: they will only propagate to the network limits
network global synchronization. and then disappear.
In the following of this paper, for the sake of simplicity, For a given Mx series, the first symbol received by a
instantaneous processing (tp=TIMEGUARD=0) and therefore terminal has followed the shortest path between the two
instantaneous reemission are assumed, consequently tT = τ• terminals. Other identical symbols, following longer paths,
This does not reflect reality. may be received with various delays. However, the
measurement by A of the time interval between the emission
of M1 and the reception of the first M2 corresponds to twice the
C. Communication protocol minimum propagation time (i.e. 2T) between A and B.
Similarly, the measurement by B of the time interval
Let us consider that terminal A needs to communicate with between the emission of M2 and the reception of the first M3
terminal B. In a first step, A and B determine the length of the corresponds again to twice the same minimum propagation
shortest path between them. For that purpose, three symbols time. At this stage, the minimum propagation time T may be
labelled M1, M2 and M3 will be used. At the beginning of a calculated by A and B. T is related to the number N of
round, terminal A sends a broadcast symbol M1 to all nodes. microrepeaters involved in the shortest path by (2):
Another broadcast symbol labelled M2 is emitted by B in T = τ (N+1). (2)
response to the reception of the first received M1 symbol. This

IEEE Communications Society 3541 0-7803-8533-0/04/$20.00 (c) 2004 IEEE


start an identical symbol (i.e. same x) after a delay τ. More simply,
the last symbol of a stream of identical symbols is not
reemitted. Moreover, retransmission is not done towards the
neighbours having initially transmitted the above symbols to
the considered microrepeater. Finally, during this phase, each
no microrepeater will only "listen" to its neighbours having
M1 ? transmitted the very first Ex symbol of all streams. This rule is
essential in the case of networks exhibiting isolated obstacles.
Compared to Mx broadcast symbols, the fundamental
yes property of Ex symbols is that they propagate from a terminal
over a limited distance only; this distance depends upon the
M1 length of the stream of Ex symbols initially emitted by a
terminal. This is illustrated in Fig. 4 for two streams of 5 and 4
symbols respectively. The use of two symbols (E0 and E1)
permits to emit adjacent frames, without idle time intervalls
between them (see Fig. 4).
no
M2 ? The following results from this procedure:
th
- to reach the n repeater, a stream of n symbols is needed,
th
yes - the time interval Tp to reach the n repeater is given by

M2 Tp = (2n-1) τ. (3)

To activate the microrepeaters, the terminals will


synchronously emit streams of Ex symbols by pair, alternating
no streams of E0 and E1, each last symbol of a stream propagating
M3 ? along a fraction of the shortest path only. The sum of the
propagation distances of a pair will be equal to the spatial
shortest path (T/τ).d, that is (N+1).d according to (2). To
yes th
activate the n repeater, the lengths of the two streams will be
respectively n and (N-n+1). Fig. 4 illustrates the activation of a
M3 microrepeater at distance 4d of a terminal.
N stream pairs are necessary to activate the N repeaters
involved. The condition of self-activation of a node is that at
Figure 3: basic transmission protocol of M symbols for microrepeaters. any time during this phase, the node will simultaneously
Emissions are identical in all directions. receive at least two (it may receive at most four) identical Ex
symbols.
Once 2T has been measured, N can therefore be calculated The fastest procedure consists in starting Ex emission
using (2). This equation illustrates a key point of the protocol immediately after terminal B received the first M3 symbol. B
that is the relationship between propagation time and distance, will therefore emit the N streams one after another. Terminal
and therefore between propagation time and number of A will emit the N streams with a delay Γ after the emission of
microrepeaters involved. This is of course a consequence of M 3.
the periodical structure of the meshed network. The last phase deals with the exchange of time series data
The next task for terminals A and B is to "activate" the N between A and B. This is now true information, no more
microrepeaters participating to the propagation along this related to the establishment of the transmission link. Our
shortest route. It is worth noting that this path may not be proposed transmission control scheme is as follows. A third set
unique, i.e. that more than N nodes may be involved. of symbols, labelled D, is used. At this stage, emitted symbols
However, if more several routes do exist, all have the same are marked for the previously selected microrepeaters only; all
length and consequently even if multipath propagation takes other nodes discard them. More precisely, the selected nodes
place there is no delay–spread and time series data from will receive and repeat these symbols, while their closest
different paths are synchronised at the receiver end. neighbours will received the symbols but not repeat them, and
To activate the N microrepeaters, a new set of symbols is the other nodes will not even receive the D symbols: a
used; these symbols are labelled E0 and E1. The transmission multipath-free channel is then established at the physical
protocol relative to these two symbols for a microrepeater is network level, with the added advantage of corresponding to
straightforward: a given Ex symbol is retransmitted by a the shortest route therefore reducing minimizing total
microrepeater only if its reception has been followed by that of propagation time.

IEEE Communications Society 3542 0-7803-8533-0/04/$20.00 (c) 2004 IEEE


rectangular mesh represented in Fig. 1 is probably not the
terminal microrepeaters optimum solution considering multiple propagations and
1 2 3 4 5 therefore unnecessary activations. A honeycomb pattern is
probably more efficient. This performance assesment is part of
0 the future work considered by the authors.
1 Another obvious issue is related with the ability of the
protocol to support many independent point-to-point flows, i.e.
2 to allow nodes to handle multiple path discoveries
simultaneously, therefore allowing fair bandwidth allocation.
3 Provided that synchronization is maintained during the totality
4 of the communication (that is the three phases plus the reset
step) two or more simultaneously active channels may coexist
5 and use the same microrepeaters on condition that
6 communications have been initiated at instants t0 and t'0 with t0
• t'0 ± nτ, n being an integer. This shift in time would allow the
7 repeaters to discriminate between different traffics.
8 More complex functions (i.e. point to multipoint
communication, terminal identification, terminal
9 supervision...) are out of the scope of this paper as they are
related to higher-level network functions.
10
11 V. CONCLUSION
12
In this paper, a methodology for routing communications
13 between two terminals has been proposed. This methodology
is also designed to save power in the network nodes first by
minimizing the amount of microrepeaters needed for
time (τ) establishing a link between two terminals (therefore saving the
energy of those nodes not required for the communication
Figure 4: partial chronograms for the emission/reception of E1 (black squares)
and E0 (open square) symbols. Instantaneous processing takes place (tp=0). canal), then by minimizing computational load at the node
Microrepeaters are at distances d, 2d, 3d… from terminal. Labels of the level (nodes can be regarded as elementary automata), and
vertical axis correspond to an emission when standing on right, and to a finally by eliminating the need for direct coordination between
reception when on left. The microrepeater at distance 4d from terminal is
nodes. This is also achieved by relying upon the large
activated at times 12 τ. Only one terminal is shown.
computing and energetic power available in the terminal units.
Moreover, the above protocol exhibits high fault tolerance.
If on-off-keying encoding is used, only two symbols D0 and The hardware of the network discussed in this paper is very
D1 may be used. A node receiving a D symbol from one or different from that of a classical data network because the
more neighbour nodes repeats it to its other neighbours. As nodes have very limited processing and storage capabilities
stated in section II.B, for a network topology as that of Fig. 1, together with small energetic budgets. Consequently, classic
a microrepeater may have to process up to four simultaneously techniques including those used in Wireless Local Area
incoming symbols, and may need to emit simultaneously up to Networks and Personal Area Networks for routing
four symbols. This property is used during this last phase to communications and for dealing with congestion and link
permit full-duplex transmission. failure, are not appropriate.
To end up the communication, terminal A broadcasts a On the other hand, while our network share strong
reset symbol, which propagation is that of a Mx symbol: after similarities together with sensor networks (limited power,
completion of the propagation, all nodes are back in an storage and processing) and ad hoc networks (multi-hop
identical state, ready for a new sequence as the one depicted in propagation of information) strong differences (possible wired
Fig. 3. A and B may now be plugged in new positions. channels, fixed topological location of nodes) make it distinct.
This reset procedure could also occur in the presence of Looking ahead, following questions are to be explored:
increasingly rampant intermittent (or total) failures to remap - what are the performances of the above algorithms
communication process. with respect to energy efficiency, latency...
- is it possible to adapt the above algorithms to
IV. PERFORMANCES microrepeaters randomly distributed.

While the above protocol needs to be tested vs. various


network geometries, it is worth mentioning that the hardware
itself, i.e. the mesh structure, also needs optimization. The

IEEE Communications Society 3543 0-7803-8533-0/04/$20.00 (c) 2004 IEEE


ACKNOWLEDGMENT

The authors would like to thank J. Collet for all the


discussions related to this work.

REFERENCES
[1] B. Warneke, M. Last, B. Liebowitz, K. Pister, "Smart Dust:
Communicating with a Cubic-Millimeter Computer", Computer, vol. 34,
n°1, IEEE Computer Society, 2001.
[2] Jason Hill, Robert Szewczyk, Alec Woo, Seth Hollar, David E. Culler,
Kristofer S. J. Pister "System Architecture Directions for Networked
Sensors", Proc. 9th International Conference on Architectural Support for
Programming Languages and Operating Systems, Cambridge (MA), USA,
2000, ACM Press, New York, pp. 93-104.
[3] Alec Woo, David E. Culler, "A Transmission Control Scheme for Media
Access in Sensor Networks", 7th Annual International Conference on
Mobile Computing and Network (Mobicom 2001), Rome, Italy, 2001.
[4] D. Marculescu et al., "Electronic Textiles: a Platform for Pervasive
Computing", Proceedings of the IEEE, vol. 91, NO. 12, December 2003.
[5] L.A. Samuelson, F.F. Bruno, J. Kumar, R.A. Gaudiana, P.M. Wormser,
"Conformal Solar Cells for the Soldiers", International Interactive Textiles
for the Warrior Conference, Cambridge, MA, USA (2002).
[6] A. Chandrakasan, R. Min, M. Bhardwaj, S-H. Cho, A. Wang, "Power
Aware Wireless Microsensor Systems," ESSCIRC, Firenze, Italy, 2002.
[7] J-M. Dilhac, D. Dragomirescu, " Configurable Communication Channel
Using Massively Distributed Wireless Microdevice Network," in
Proceedings of the XVIII World Telecommunication Congress, Paris –
France, 2002.
[8] J-M. Dilhac, D. Dragomirescu, "In-Building Communication using
Massively Distributed Wireless Microdevice Network", IEEE Consumer
Communication and Networking Conference, Las Vegas, USA, January
2004.

IEEE Communications Society 3544 0-7803-8533-0/04/$20.00 (c) 2004 IEEE

You might also like