14-05-2013, 03:21 PM
Simulation-based optimization of communication protocols for large-scale wireless sensor networks
[attachment=53883]
Abstract
The design of reliable, dynamic, fault-tolerant
services in wireless sensor networks is a big challenge and a
hot research topic. In this paper an optimization method is
proposed that can be used to tune parameters of the
middleware services and applications to provide optimal
performance. The optimization method is based on
simulation, and is capable of handling ‘noisy’ error surfaces.
The proposed optimization algorithm is illustrated by a new
spanning-tree formation algorithm, which can effectively
operate even if links between nodes are asymmetrical.
INTRODUCTION
In the near future large-scale sensor networks will be the key
elements of embedded systems used in space and aviationrelated
challenges, e.g. monitoring and control of safety
critical systems [1], Smart Surfaces, Smart Dust [2], or can
be used to make everyday life more comfortable, e.g.
Intelligent Spaces [3]. These sensor networks often use
distributed operating system-like services (called middleware)
over wireless communication protocols, which must
be fault tolerant and adaptive because of the dynamic
network topology and changing mission objectives. The
design of such middleware services is not straightforward,
since the sensors have limited resources, and thus the used
protocols are usually very simple compared to ones used in
wired communication schemes. The nondeterministic nature
of the environment is another factor making the design more
difficult. This paper presents a simulation-based
optimization method that can be used to tune the algorithms
used in the middleware layer. Also some results are
presented that were gained by the proposed method.
THE TARGET SYSTEM
A very successful, low-cost prototype field-node (mote)
family was developed at Berkeley. The used variant (MICA)
of the Berkeley motes (see Figure 1) includes an 8-bit, 4
MHz Atmel ATMEGA103 microcontroller, 128kB program
memory, 4KB RAM, and an RFM TR1000 radio chip
capable of providing 50 kbit/s transmission rate at 916.5
MHz. The motes can also accommodate a set of
interchangeable sensors (temperature, light, magneto, sound,
etc.) [4].
The motes use a small operating system called TinyOS,
designed to provide the necessary services in despite of the
very limited hardware resources. It contains a complete
network stack with bit-level error correction, medium access
layer, network messaging layer, and timing [5].
WIRELESS NETWORK SIMULATOR
The probabilistic wireless network simulator (Prowler) is an
event-driven simulator that can be set to operate in either
deterministic mode (to produce replicable results while
testing the application) or in probabilistic mode (to simulate
the nondeterministic nature of the communication channel
and the low-level communication protocol of the motes). It
can incorporate arbitrary number of motes, on arbitrary
(possibly dynamic) topology, and it was designed so that it
can easily be embedded into optimization algorithms. The
simulator runs under MATLAB, thus it provides a fast and
easy way to prototype applications, and has nice
visualization capabilities. The graphical user interface of
Prowler is shown in Figure 2.
OPTIMIZATION FRAMEWORK
The network simulator can be used to test protocols and
algorithms and it also can provide metrics on the
performance of the tested application. Similarly to the core
of the simulator, the applications can be parameterized, so
different settings can easily be tested. The proposed
optimization algorithm is built around the simulator and it
calls the simulator with the required parameters.
In the development phase of new protocols, a typical
problem is to provide optimal performance in some metric,
versus a certain set of design parameters. This is a simple
optimization problem leading to the search of an error
surface above a parameter space. There are multiple
methods for solving this problem, if the error surface is well
defined. The main idea behind these methods is some kind
of exploration of the error surface, either a gradient-based
method, Monte-Carlo search, or an annealing method [8].
These optimization methods use so-called ‘function calls’ to
compute the value of the cost function. The more
computationally expensive the function call the more
important it is to keep the number of function calls low.
CONCLUSIONS
An optimization method was proposed, that is able to handle
noisy error surfaces. This method, combined with a
probabilistic wireless network simulator, can be used to
optimize parameters of middleware services and
applications in wireless sensor networks.
[attachment=53883]
Abstract
The design of reliable, dynamic, fault-tolerant
services in wireless sensor networks is a big challenge and a
hot research topic. In this paper an optimization method is
proposed that can be used to tune parameters of the
middleware services and applications to provide optimal
performance. The optimization method is based on
simulation, and is capable of handling ‘noisy’ error surfaces.
The proposed optimization algorithm is illustrated by a new
spanning-tree formation algorithm, which can effectively
operate even if links between nodes are asymmetrical.
INTRODUCTION
In the near future large-scale sensor networks will be the key
elements of embedded systems used in space and aviationrelated
challenges, e.g. monitoring and control of safety
critical systems [1], Smart Surfaces, Smart Dust [2], or can
be used to make everyday life more comfortable, e.g.
Intelligent Spaces [3]. These sensor networks often use
distributed operating system-like services (called middleware)
over wireless communication protocols, which must
be fault tolerant and adaptive because of the dynamic
network topology and changing mission objectives. The
design of such middleware services is not straightforward,
since the sensors have limited resources, and thus the used
protocols are usually very simple compared to ones used in
wired communication schemes. The nondeterministic nature
of the environment is another factor making the design more
difficult. This paper presents a simulation-based
optimization method that can be used to tune the algorithms
used in the middleware layer. Also some results are
presented that were gained by the proposed method.
THE TARGET SYSTEM
A very successful, low-cost prototype field-node (mote)
family was developed at Berkeley. The used variant (MICA)
of the Berkeley motes (see Figure 1) includes an 8-bit, 4
MHz Atmel ATMEGA103 microcontroller, 128kB program
memory, 4KB RAM, and an RFM TR1000 radio chip
capable of providing 50 kbit/s transmission rate at 916.5
MHz. The motes can also accommodate a set of
interchangeable sensors (temperature, light, magneto, sound,
etc.) [4].
The motes use a small operating system called TinyOS,
designed to provide the necessary services in despite of the
very limited hardware resources. It contains a complete
network stack with bit-level error correction, medium access
layer, network messaging layer, and timing [5].
WIRELESS NETWORK SIMULATOR
The probabilistic wireless network simulator (Prowler) is an
event-driven simulator that can be set to operate in either
deterministic mode (to produce replicable results while
testing the application) or in probabilistic mode (to simulate
the nondeterministic nature of the communication channel
and the low-level communication protocol of the motes). It
can incorporate arbitrary number of motes, on arbitrary
(possibly dynamic) topology, and it was designed so that it
can easily be embedded into optimization algorithms. The
simulator runs under MATLAB, thus it provides a fast and
easy way to prototype applications, and has nice
visualization capabilities. The graphical user interface of
Prowler is shown in Figure 2.
OPTIMIZATION FRAMEWORK
The network simulator can be used to test protocols and
algorithms and it also can provide metrics on the
performance of the tested application. Similarly to the core
of the simulator, the applications can be parameterized, so
different settings can easily be tested. The proposed
optimization algorithm is built around the simulator and it
calls the simulator with the required parameters.
In the development phase of new protocols, a typical
problem is to provide optimal performance in some metric,
versus a certain set of design parameters. This is a simple
optimization problem leading to the search of an error
surface above a parameter space. There are multiple
methods for solving this problem, if the error surface is well
defined. The main idea behind these methods is some kind
of exploration of the error surface, either a gradient-based
method, Monte-Carlo search, or an annealing method [8].
These optimization methods use so-called ‘function calls’ to
compute the value of the cost function. The more
computationally expensive the function call the more
important it is to keep the number of function calls low.
CONCLUSIONS
An optimization method was proposed, that is able to handle
noisy error surfaces. This method, combined with a
probabilistic wireless network simulator, can be used to
optimize parameters of middleware services and
applications in wireless sensor networks.