End-to-end Routing for Dual-Radio Sensor
Networks
Thanos Stathopoulos
†
Martin Lukac
‡
Dustin McIntire
♯
John Heidemann
⋄
Deborah Estrin
‡
William J. Kaiser
♯
† Institute of Computer Science (FORTH–ICS)
Foundation for Research and Technology–Hellas
GR-711 01, Heraklion, Crete, Greece
Center for Embedded Networked Sensing
‡ UCLA, Department of Computer Science
♯ UCLA, Department of Electrical Engineering
⋄ USC, Information Sciences Institute
{thanos@forth.ics.gr, mlukac@lecs.cs.ucla.edu, dustin@seas.ucla.edu, johnh@isi.edu, destrin@cs.ucla.edu, kaiser@ee.ucla.edu}
Abstract— Dual-radio, dual-processor nodes are an emerging
class of Wireless Sensor Network devices that provide both low-
energy operation as well as substantially increased computational
performance and communication bandwidth for applications.
In such systems, the secondary radio and processor operates
with sufficiently low power that it may remain always vigilant,
while the the main processor and primary, high-bandwidth radio
remain off until triggered by the application. By exploiting
the high energy efficiency of the main processor and primary
radio along with proper usage, net operating energy benefits
are enabled for applications. The secondary radio provides
a constantly available multi-hop network, while paths in the
primary network exist only when required. This paper describes a
topology control mechanism for establishing an end-to-end path
in a network of dual-radio nodes using the secondary radios
as a control channel to selectively wake up nodes along the
required end-to-end path. Using numerical models as well as
testbed experimentation, we show that our proposed mechanism
provides significant energy savings of more than 60% compared
to alternative approaches, and that it incurs only moderately
greater application latency.
I. INTRODUCTION
Ever-increasing application demands in conjunction with
advances in low-power hardware design have resulted in an
increasing use of larger, more powerful sensor nodes [1], [2].
In addition to a 32-bit CPU, those nodes include sophisticated
peripherals and megabytes of RAM and flash as well as a high-
bandwidth 802.11 radio. 32-bit nodes are used in standalone
wireless sensor network deployments [3], [4], [5] as well as
in tiered architectures, where they operate in conjunction with
microcontroller-based WSN devices [6], [7], [8].
When the 32-bit nodes are used in a tiered architecture, they
must communicate with the network of microcontroller-based
nodes (typically 8- or 16-bit motes [9], [10]). For this reason
a new generation of 32-bit nodes as for example the LEAP
node [2] include an on-board low-power microcontroller (for
constantly vigilant operation) and a second, low-bandwidth
radio. As a result those nodes enable not just tiered computing
but also tiered radio networking. When a node has multiple
radios with different communication capabilities and power
properties, the question becomes: How should such a multi-
radio system be applied to best benefit energy and application
demands? It is important to note that the high-bandwidth radio
operates with much greater energy efficiency than the low-
bandwidth radio, in terms of energy per bit transmitted (for
example, 112 nJ/bit for 802.11g as opposed to 979 nJ/bit for
802.15.4 [2]). However, the larger radio also has a much higher
state transition cost and idle energy consumption, more than
10 times that of the low bandwidth radio [1], [2], [11]. It is
therefore counter-productive to use the high-bandwidth radio if
there is little or no data to send or if data needs to be sent only
occasionally. Instead, in order to reduce energy consumption,
the high-bandwidth radio should be kept off, to be activated
only when there is a significant amount of data that needs to
be transmitted. The low-bandwidth radio, on the other hand, is
less energy efficient but consumes much less energy when idle
and is able to quickly transition from sleep to active state, send
the necessary data, and then deactivate. It is therefore ideal
for transmitting small amounts of data as well as remaining
“vigilant” for long time periods, especially when techniques
such as Low-Power Listening [12] are used.
In a multihop network of nodes that maintain their main
CPU and high-bandwidth radio in a low operating duty cycle
so as to conserve energy, end-to-end paths do not always exist.
In sensor network applications where observing a phenomenon
for which a well-established model exists, this problem can be
solved by either a static or an adaptive scheduling algorithm,
where nodes coordinate in order to guarantee that their wakeup
times are synchronized and to perform other low duty cycle
coordination functions. However, when such a model does not
exist and may not be learned as in seismic event detection,
or when timely notification of an event is required, as in
- 1
- 2
- 3
前往页