# RFID Middleware Design: Optimal Scheduling RFID Reader Networks

Document Sample

```					RFID Middleware Design: Optimal Scheduling RFID
Reader Networks Based on Swarm Intelligence

Hanning Chen
October 28nd, 2006
Outline
 Introduction
 A brief review of PSO and B- PSO
 RFID Readers Scheduling and GPP
 Optimal Scheduling for RFID Reads
networks
 Conclusions
Introduction
 RFID middleware design
 Scheduling Problem of RFID reader
networks
 construction of GPP using   evolutionary
algorithm
 Our method
Particle Swarm Optimization (PSO)
 Particle Swarm Optimization (PSO) applies to
concept of social interaction to problem solving.
 It was developed in 1995 by James Kennedy
and Russ Eberhart [Kennedy, J. and Eberhart, R. (1995).
“Particle Swarm Optimization”, Proceedings of the 1995 IEEE
International Conference on Neural Networks, pp. 1942-1948, IEEE
P          r          e         s          s         .         ]
 It has been applied successfully to a wide
variety of search and optimization problems.
 In PSO, a swarm of n individuals communicate
either directly or indirectly with one another
 PSO is a simple but powerful search technique.
PSO Velocity Update Equations

vnew
id      wi  v
old
id     c1  rand1  ( pid  xid )  c2  rand 2  ( pgd  xid )
xid  xid  vid
new  old    new

 Given a collection of RFID readers laid out in some manner, we
can construct the associated conflicting graph G = (V,E) where
each vertex v ∈ V corresponds to a RFID reader and each edge e ∈
E indicates that those two sensors can be operated in parallel. In
other words there are no constraints between these two readers. For
example, the conflicting graph corresponding to the RFID reader
layout of       Figure a is given in Figure b.
 Readers in any given partition of the conflicts graph can read
simultaneously without interference. Thus it makes sense to fire
every reader in a partition when firing one reader in the partition.
 Now the optimal schedule can be determined by finding the
maximum partition and partitioning the graph into partitions.
networks
（1） Particle representation
In our work the direct encoding scheme is applied to encode the individuals.
The dimension of each particle is set as equal to the number of sensor
reader “N”. Each element in the dimension is corresponding to the absence
of particular readers, whose entries can only be “0” or “1’’. A bit “0” in an
individual indicated the absence of the corresponding reads. Otherwise a bit
“1” in an individual indicated the presence of the corresponding reads. For
example, a particle’s current position is “001101”. It denotes the 6 reads in
our system and “1” implies presence of that particular sensor in the clique
which the particle is representing.
（2）Initialization
Initially M individuals forming the population should be randomly
generated and each consists of N parameters. These individuals may be
regard as particles in terms of PSO. In addition, the learning parameters,
such as and , inertia weight should be assigned in advance.
networks
(3) Fitness function design
To evaluate the performance of an individual, a predefined fitness function should be
formulated. The fitness function takes into account four parameters:
The f is calculated as the reciprocal of C as follows:

Where N is number of sensors, ‘T’ is the transaction time of the partition, ‘W’ is the
weight attached to this group of readers.                  are the weights given to each
one of them and the importance of each one of them differed.
The transaction time for a partition can be calculated as

Where is the transaction time of the ith member (reader) that forming the partition.C
is the summation of all the possible conflicts that the members of the clique have with
the nodes still remaining in the graph to be partitioned.It should be noted that the four
parameters in cost function should be normalized this normalization is done after
merging the pbest and the present vectors together.
Optimal Scheduling for RFID Readers networks

(4) Update dependencies and transaction time
The velocity and position are updated according to
Eqs above. After this step the individuals associated
with both the dependencies and transactions times
are updated to produce new best-performing
i   n    d     i  v     i   d    u    a    l   s    .

(5) Termination condition
The proposed algorithm is performed until the
Fitness is small enough, or a pre-determined
number of epochs is passed. It is expected that,
after a certain number of iterations, all the reader
will grouped and the optimal group can be obtained.
Pseudocode for implementing our
algorithm
Begin;
Generate random population of N particles, i.e. the initial transaction times
and conflicts should be given;
For each individual i=1: N
calculate fitness value ();
end
For each particle i= 1: N;
Set pBest as the best position of particle i;
If fitness value () is better than pBest;
pBest(i)=f(i);
End;
Set gBest as the best fitness of all particles;
For each particle;
Calculate particle velocity and position according to Eqs.(1-4);
End;
Check if termination is true;
End
Conclusions and Future Work

This paper is devoted to giving a new strategy for optimal
scheduling of RFID read networks. A swarm intelligence based
algorithm, binary particle swarm optimization is employed to search
through space for an optimization problem.

In the future work, some improved swarm intelligence based
algorithm or artifical life methodology can be incorporated to solve
the problem of optimal scheduling of RFID read networks. By this
way, the robust and powerful function of RFID middleware can be
achieved. The insights presented in this paper will be certainly
found to be useful in our RFID Lab. In fact the experiment
environment has been setup and some primary results will be given.
Due to the limit of the conference date all those will be done in our
f     u     t      u       r    e          w      o      r    k     .
Thanks
Email: chenhanning@sia.cn
ADDRESS: Shenyang Institute of Automation, Chinese