U.S. patent application number 11/200937 was filed with the patent office on 2007-02-15 for design and construction of wireless systems.
Invention is credited to David Abusch-Magder.
Application Number | 20070037580 11/200937 |
Document ID | / |
Family ID | 37743183 |
Filed Date | 2007-02-15 |
United States Patent
Application |
20070037580 |
Kind Code |
A1 |
Abusch-Magder; David |
February 15, 2007 |
DESIGN AND CONSTRUCTION OF WIRELESS SYSTEMS
Abstract
In the design and/or construction of a wireless communication
system, a parameter optimization determination is performed. Such
determination is accomplished by first clustering parameters with
strong interdependence. In a second operations a subset of the
clustered parameters are removed from consideration by imposing
constraints on the system and/or on the parameters to yield a
similar system determination problem. In a third operation,
parameters remaining after the second operation are optimized for a
chosen performance metric. Finally, parameters removed in the
second operation are optimized for a chosen performance metric with
the parameters optimized in the third operation, set to an
optimized value determined in the third operation.
Inventors: |
Abusch-Magder; David;
(Maplewood, NJ) |
Correspondence
Address: |
Bruce S. Schneider
1153 Long Hill Road
Stirling
NJ
07980-1007
US
|
Family ID: |
37743183 |
Appl. No.: |
11/200937 |
Filed: |
August 10, 2005 |
Current U.S.
Class: |
455/446 |
Current CPC
Class: |
H04W 16/22 20130101 |
Class at
Publication: |
455/446 |
International
Class: |
H04Q 7/20 20060101
H04Q007/20 |
Claims
1. A method for the design or construction of a wireless
communication system having a plurality of antennae including the
step of optimizing a set of parameters of said system wherein said
system has original imposed constraints, said optimization step
comprising the operations of 1) clustering said parameters into
groups of related parameters, such that two parameters may belong
to the same group only if their joint effect on a performance
metric according to a suitable measure equals or exceeds a
threshold; 2) imposing one or more additional constraints with the
effect that: a) in each of at least some groups, at least one
parameter is effectively fixed; and b) the function of signal
strength versus geographic location in relation to the
corresponding function without additional constraint satisfies a
similarity criterion; 3) optimizing under the combined original and
additional constraints relative to a first performance metric of
the wireless communication system to determine a set of one or more
optimized values for each parameter; and 4) under the original
constraints, optimizing at least a portion of the previously fixed
parameters relative to a second performance metric of the wireless
communication system, wherein those parameters optimized in
operation 3 are constrained to said sets of optimized values.
2. The method of claim 1 wherein said first performance metric and
said second performance metric are the same.
3. The method of claim 1 wherein the spatial orientation of said
antennae comprises at least a subset of said parameters.
4. The method of claim 1 wherein the geographic location of said
antennae comprises a subset of said parameters.
5. The method of claim 1 wherein said first performance metric
comprises signal strength.
6. The method of claim 1 wherein said first performance metric
comprises network capacity.
7. The method of claim 1 wherein said first performance metric
comprises network coverage.
8. The method of claim 1 wherein at least one parameter in at least
5% of said groups is removed in operation 2.
9. The method of claim 1 wherein at least one of said parameters is
removed from at least 50% of said groups in operation 2.
10. The method of claim 1 wherein at least one of said parameters
is removed from at least 90% of said groups-in operation 2.
11. The method of claim 1 wherein said additional constraint lacks
a physical analogue.
12. The method of claim 1 wherein said parameters optimized in
operation 3 or 4 comprise antenna tilt or antenna azimuth.
13. The method of claim 12 wherein said parameter optimized in
operation 3 or 4 comprises antenna site location.
14. The method of claim 1 wherein said parameter optimized in
operation 3 or 4 comprises antenna site location.
15. The method of claim 1 wherein said parameters include discrete
parameters.
16. The method of claim 1 wherein said parameters include
continuous parameters.
17. The method of claim 1 wherein said parameters include carrier
frequency.
18. The method of claim 1 wherein said parameters include antenna
type.
19. The method of claim 1 wherein said plurality of optimized
values comprise a range of optimized values.
20. The method of claim 1 wherein said additional constraint
comprises a change in antenna radiation pattern.
21. The method of claim 1 wherein each pair of parameters of a
group has a coupling of at least 0.01.
22. The method of claim 21 wherein said coupling is at least 0.1.
Description
TECHNICAL FIELD
[0001] This invention relates to wireless communication systems and
more particularly to systems employing a multiplicity of base
stations.
BACKGROUND OF THE INVENTION
[0002] For wireless communication systems a geographic area is
divided into a plurality of cells, generally with at least one base
station in each cell. A mobile communication unit wishing to
transmit information, e.g. voice or data, establishes a
communication link to one or more base stations. (Generally, but
not invariably, the contacted base stations are located spatially
near the mobile unit.) The information is transmitted from the
mobile unit to the base station and then from the base station into
the communication network. Although requirements vary depending on
the technology and bandwidth, acceptable operation of the system
for voice calls typically requires signal strength of approximately
-95 dBm or greater.
[0003] Numerous system parameters affect the quality of system
operation. For example, the number of base stations, the geographic
location of these base stations, the type and number of antennae at
each base station, the orientation of each antenna (azimuthal angle
1 FIG. 1, and radial angle 2 in FIG. 1 relative to projection 3),
transmit power, pilot power, carrier frequency and antenna type all
typically influence system efficacy. (Such parameters include both
discrete parameters, e.g. antenna type, and continuous parameters,
e.g. antenna orientation parameters.) Additionally, there are
generally constraints on these parameters and on the system. (In
the context of this description, both types of constraints are
subsumed by the term constraints.) Constraints on parameters are
imposed by physical and practical considerations. As exemplary of
the former, the mounting brackets on antennae often physically
preclude some portion of the azimuthal and radial angular spectrum.
Similarly, topography such as hills and/or structures such as tall
buildings make antenna azimuthal and radial angles directed toward
such natural and man-made obstacles impractical. System constraints
are exemplified by antenna radiation patterns, a variety of
technology requirements some of which are imposed by a previously
agreed upon standards, and the signal processing design of mobile
terminals and base stations. Thus, in determining the potential
quality of a communication system the parameters should be
considered in view of these constraints.
[0004] Since a plethora of parameters and constraints influence the
system, and since generally there is a strong interdependence
between and among parameters, any attempt to find an optimal design
and implement such a system involves an extremely complicated
calculation dependent on these parameters and constraints. Indeed,
the extent of interdependence in a wireless system environment
makes the problem particularly difficult. Therefore the search for
optimal designs has not been carried out directly, but instead has
advanced through the search for superior if suboptimal designs
based on a variety of tractable algorithms.
[0005] A measure of effectiveness of such an algorithm is the
improvement in calculated system performance (from the inception of
calculations to the conclusion) per unit time of calculation. One
of the most commonly employed approximations for satisfying the
improvement per unit time criterion involves an iterative approach
to optimization. Two or more specialized algorithms each of which
optimizes a subset of the parameters is employed iteratively to
satisfy the improvement/unit time criteria. (Optimized parameters
in the context of this disclosure is not necessarily the best
choice of parameters possible, but a choice of parameters that
improves the calculated performance of the system.) For example, it
is often necessary for the design and construction of a system to
perform a calculation to choose a number of base locations from an
ensemble of possibilities, e.g., choose 70 locations for base
station placement from 100 possibilities, and to set the azimuthal
and radial orientation for each antenna at each chosen location. To
make the problem manageable, in one approach, an iterative
procedure is employed at each cycle of the iterations including, as
applied to the earlier example, A) starting with all possible sites
and optimizing the antennae orientations or choosing a starting
orientation for the antennae B) optimizing by a suitable algorithm
to eliminate locations e.g. 5 locations, and then C) optimizing by
a suitable algorithm angles for the antennae (e.g. 3 antennae), at
each remaining location. (The actual number of antennae at each
location as previously discussed is a system constraint.) Thus, as
an illustration a flow chart for the example (pick 70 locations
from 100 possible locations and set the orientation for the three
antennae at each of the 70 chosen locations) is shown in FIG. 2. In
the first step 21, the antennae angles are initialized, for
example, to be the allowed value closest to pointing out to the
horizon with equal spacing in the azimuthal direction for the 300
possible antennae (3 at each of 100 possible locations). An antenna
optimization algorithm is then utilized to choose desirable
orientations for each of the antennae. Then in step 22 a location
optimization is used to eliminate a number of locations, e.g. 5.
(Of course fewer locations could be eliminated in such step but at
a cost of substantially increased computing time or more locations
could be eliminated at the cost of decreased performance
improvement. The number of locations actually eliminated at each
step is a matter of choice depending on the system design and
construction goals.) The antennae angle optimization algorithm in
step 23 is then used to choose a desirable orientation for each of
the remaining 285 antennae distributed at the remaining 95
locations. Fixing the antennae orientation determined in step 23
the procedure then cycles with the location optimization algorithm
in step 24, choosing 90 of the remaining 95 locations. (It is
necessary to perform the antennae optimization for desirable
improvement after each elimination step due to the strong
interdependence of location and antennae orientation parameters as
discussed earlier.) In the continuing iterative procedure the
orientation of the remaining 270 antennae are optimized in step 25.
Alternation continues until 70 locations with a total of 210
optimized antennae orientations are chosen. In all, 6 location
optimizations are performed on respectively 100, 95, 90, 85, 80,
and 75 locations as well as 7 antennae optimization procedures on
respectively 300, 285, 270, 255, 240, 225 and 210 antennae.
[0006] Thus despite the simplification of fixing parameters or
optimizing over a subset of variables, the problem still requires
multiple optimization steps, each of 25 which is a difficult
optimization of a large number of parameters, e.g. antennae
orientation parameters and antennae locations. Accordingly,
although simplification is achieved, the approach is nevertheless
not simple. Substantial time (generally 1 week or more) even on a
high speed computer is still needed to obtain an accurate solution
for even modest system design problems. Decreasing the required
time for obtaining an accurate solution provides many advantages.
For example, it is possible to perform the solution process
repeatedly in an acceptable time period to accommodate the
evolution of design goals and constraints inherent in reaching a
design that satisfies the various evolving exigencies envisioned
for the system. Additionally, the ability to reach a rapid solution
aids in the adjustments and engineering choices involved in
constructing a system. As a result both the design and construction
of wireless communication systems would benefit from an approach
that enhances the performance improvement per unit calculation time
for determinations that are part of the design and/or construction
process.
SUMMARY OF THE INVENTION
[0007] It has been found that during optimization to obtain
parameters for a design or construction process, the speed of
determining accurate parameters for use is significantly enhanced
by employing specific expedients. This determination step expedient
is particularly useful in problems having a strong coupling between
at least some of its parameters. (Strong coupling between two
parameters in the context of this invention means a coupling
between parameters x and y defined as a ratio of is derivatives (or
their discrete analog) of at least 0.01 as to the underlying
performance metric (e.g. system coverage capacity, latency, maximum
throughput, maximum bandwidth, total revenue generated or some
combination thereof)) for which at least one optimization operation
is to be performed. In particular the coupling is the maximum value
obtained over all allowed parameters, x.sub.0, y.sub.0 (considering
the constraints) of the performance metric, f, of the following
ratio: f .function. ( x 0 , y 0 ) .times. .differential. 2 .times.
f .differential. x .times. .differential. y .times. ( x 0 , y 0 )
.differential. f .differential. x .times. ( x 0 , y 0 ) .times.
.differential. f .differential. y .times. ( x 0 , y 0 )
##EQU1##
[0008] The parameter determination step of the design and/or
construction process includes four operations. The first operation
is clustering the parameters being considered. That is, all the
parameters whose values are to be determined are grouped into at
least two clusters. The elements (parameters) of a cluster are
chosen so that the coupling between each possible pair of elements
in the cluster is at least 0.01, preferably at least 0.1.
[0009] In the second operation of the determination step at least
one parameter is removed from consideration in at least 5 percent
of the clusters by modifying the constraints of the original
problem. In this parameter removal the parameters of an individual
cluster are considered relative to the modified constraints. For
example, a few constraint is established by a change of the antenna
radiation pattern such that each antenna has, at any geographic
point, a propagated signal strength of the maximum obtainable at
that point from such antenna when all possible orientations under
the original constraints of that antenna are considered. By
imposing this new constraint the antennae orientation parameters of
the original problem are removed from being a parameter whose
variation affects the performance metric. In essence, the removed
parameter is fixed. However, such removal of parameters should not
be arbitrary. It should be done in a manner that satisfies a
similarity criterion, i.e. at least one value taken by the function
of signal strength versus geographic location under the original
set of constraints should, for each antenna be similar to the
signal strength under the combined original and new constraints. In
the context of the invention similar signal strength occurs when
the signal strength function for at least one set of parameters
after imposing the new constraint(s), differs by no more than 3 dB
from the signal strength function under the original constraints
for the same chosen set of parameters for some value of the frozen
parameters and for at least one point in at least 50% of 20 equal
sized, regions covering the cell in which the antenna is located.
(The cell is defined as all geographic points where the signal
strength from the base station being considered under the new
constraints and for the chosen parameter set is greater than -95
dBm.) For the previous example of eliminating antennae orientation
parameters with the new constraint (imposed by a change of the
antenna's radiation pattern), for every location in the cell the
signal strength under the combined original and new constraints is
equal to the signal strength under the original constraints for the
antennae orientation yielding the maximum signal strength at such
point. Thus the similarity criterion is satisfied by the exemplary
constraint. (Significantly, the imposed limitation in this example
is not physically realizable. The constraints imposed to remove
parameters need not comport with physically realizable or practical
system configurations or parameter values provided the similarity
criterion is satisfied.)
[0010] In the third operation of the determination step, a suitable
algorithm is employed using the new constraints to optimize all, or
a portion of, the parameters that were not removed in the second
operation. Again, in the example, the parameter optimized would be
all or a portion of the parameter governing the presence of a cell
at a specific location since the coupled parameters of antenna
orientation are removed in operation 2. It is advantageous but not
essential that the optimized parameters are those that have a
significant effect on the signal versus location for at least one
antenna. (A parameter has a significant effect on the signal
strength versus location for an antenna if the dependency of signal
strength over this parameter varies more than 10 percent over the
range allowed for at least one given choice of all other
parameters.) Since elimination of an antenna location with its
antennae certainly significantly affects (sets to zero) the signal
strength versus location function associated with such antennae
this parameter has a significant effect and thus is advantageously
optimized. is In the final operation of the determination step, the
parameters optimized in the third operation are set at an optimized
value (or constrained to a range of values) and the parameters
removed in operation 2 are optimized using the constraints of the
original problem (those before imposing the new constraints.) Thus
in the example the locations chosen in operation 3 would be used
and the antennae orientations for antennae at these chosen
locations would be optimized.
[0011] By using the procedure of the invention an accurate
determination is made in a significantly shorter time than the
typical iterative approach. This improvement is demonstrated by
considering the operations required for the exemplary problem. In
the first operation step the parameters are clustered so that each
cluster contains a location and the orientation parameters for the
antennae at that location. In the second operation the orientation
parameters are removed from consideration by imposing the new
signal strength maximum criterion. In the third operation an
optimization is done over the location parameters so, in the
example given, of the initial 100 locations, 30 are eliminated.
Then in the fourth operation the 70 optimized locations are fixed
and optimization is done over 210 antenna orientations. Most
significantly, the optimizations in the iterative process of 285
orientations for 95 locations (step 23 in FIG. 2), 270 orientations
for 90 locations (step 25), 255 orientations for 85 locations, 240
orientations for 80 locations, and 225 orientations or 75 locations
are totally eliminated. Thus a significant time saving is
possible.
BREIF DESCRIPTION OF THE DRAWINGS
[0012] FIG. 1 is illustrative of parameters involved in antenna
orientation;
[0013] FIG. 2 is a flowchart of calculations involved in the design
and construction process,
[0014] FIG. 3 exemplifies possible imposed constraints for removing
parameters; and
[0015] FIG. 4 is illustrative of concepts involved in the
similarity criterion.
DETAILED DESCRIPTION
[0016] As discussed, the invention involves a process for design
and/or construction of a wireless communication system including a
step for determining parameters used in such system design and/or
construction. The parameter determination step includes four
operations. Not all conceivable parameters defining a system need
be determined. A set of parameters (generally a subset of all
possible parameters) is chosen for determination. In the design
and/or construction process the constraints are also set. These
constraints include both system constraints and parameter
constraints. For example, chosen constraints on the system as
previously discussed include such things as antenna radiation
patterns, signal processing design of mobile terminals and base
stations, and requirements of applicable standards. Similarly,
constraints on parameters are also set. For example, acceptable
antenna orientations to avoid physical signal barriers and antenna
orientations allowed by the antenna mountings are imposed on the
parameters.
[0017] Once the parameters to be determined, the system constraints
and the parameter constraints are chosen, the parameters are
clustered into at least two clusters of at least two such
parameters each. The parameters of a cluster are chosen so that the
coupling between each possible pair of parameters in the cluster is
at least 0.01. Thus, for example, a possible cluster could include
a specific location as one parameter, and the orientation
parameters for each antenna at that location. As can be seen the
coupling between each of these parameters would be generally
greater than 0.01, and in fact greater than 0.1, because if the
location is eliminated during optimization the effect of antennae
orientation for antennae at that location would disappear.
[0018] In the second operation of the determination step in at
least 5% of the clusters at least one parameter is removed as a
variable by redefining the constraints of the original problem.
Thus additional constraint(s) (either system or parameter
constraints) are imposed so that the requisite parameters are
removed as a variable. (In the context of this invention an
additional constraint comprehends both the inclusion of new
constraint(s) and the modification of the restrictions of an
original constraint.) Although, as discussed, an advantageous
result is achieved by removing from consideration a parameter from
each of at least 5% of the clusters, it is preferable to remove one
or more parameters from at least 50%, more preferably 90% of the
clusters. Additionally, it has been found to be of further
advantage to remove at least 5%, preferably 10%, most preferably
20% of parameters in total from each cluster during the second
operation of the determination step to further reduce computational
complexity.
[0019] Although the particular constraint chosen to remove
parameters from consideration during the third operation of the
determination step is not limited to any specific form, it is often
particularly useful to take into account many or all of the various
possible values which might be assumed by the variable to be
eliminated. This should be done without regard to the achievability
(physical, economical or otherwise) of the constraint so created.
For example, one useful, imposed constraint is that each antenna
has a propagated signal strength at any geographic point that is
the maximum obtainable from such antenna. Thus for example, as
shown in FIG. 3 in a cell 31 with base station 32 the maximum
signal strength at point 34 is attained for a particular antenna
having a .theta. orientation parameter illustratively of 20 degrees
and a .phi. orientation parameter illustratively of 37 degrees.
Similarly at geographic point 35 the maximum signal strength is
illustratively obtained for a .theta. of 15 degrees and a .phi. of
58 degrees. These maximum values at each geographic point are
employed as the new constraint even though the antenna parameters
that yield such maximum values are generally different from one
geographic point to another. Thus this imposed constraint fixes the
signal strength dependency on, for example, antennae parameters,
encountered at each geographic location and accordingly removes
.theta. and .phi. of the antenna as a parameter that affects such
signal strength. The .theta. and .phi. parameter is thus not
optimized in the third operation for the determination step since
it has been set in operation 2. It should be noted that the
additional constraint just illustrated is not physically realizable
but nevertheless quite useful.
[0020] Imposition of the new constraint to remove parameters should
be done so that a similarity criterion is satisfied, i.e., so that
the problem presented in the determination step is similar both
before and after imposition of the new criterion. The determination
problem before and after imposition of new constraint(s) is similar
if the signal strength versus geographic location behavior under
the original set of constraints as modified by the new
constraint(s) for each antenna is similar to that under the
original constraints. This comparison to determine similarity is
done in the context of a set of regions that cover the cell in
which the antenna is located. The cell, in turn, is defined by the
region where the signal strength from the antenna being considered
under the original constraints as modified by the new constraints
and for the set of non-removed parameters being considered for
satisfying the similarity criterion is greater than -95 dBm. This
cell region is divided into 20 regions of equal area. Thus as shown
in FIG. 4, if the cell region 41 is considered, it is composed of
twenty regions, 42, of equal area.
[0021] For at least one such cell division into 20 equal areas, and
for at least 50% of the 20 equal-sized regions in such division
there should be at least one point within the region such that the
signal strength at this point differs by no more than 3 dB after
imposing the new constraint(s) when compared to its value under the
original constraints. (Many potential divisions of the cell are
possible but the similarity criterion need be satisfied for only
one such division. Additionally it is possible that the cell
includes discrete portions that do not have a common boundary with
other portions of the cell. Possible divisions include partitioning
such portions into the 20 regions to produce the equal area
division.) Thus as shown in FIG. 4 if the points 42 have a signal
strength that differs no more than 3 dB after imposition of the new
constraint from before such imposition, the similarity criterion is
satisfied. There well may be many more points that satisfy the 3 dB
limitation but it is sufficient that there are at least 10 regions
with at least one such point each. It should be noted that every
point within a region has a multiplicity of signal strength values
since there is one value corresponding to each possible combination
of the allowed parameter values for the constraints imposed. If one
of these possible values after imposition of the new constraints is
equal or within 3 dB of one such value at the same point and for
the same set of parameters before imposition of the new constraints
then a point in the region satisfies the 3 dB limitation. (For each
particular point comparison is made with the same set of parameters
and for any value of the parameters removed in the second step) as
discussed. However the parameter set used at one point need not be
the same as the parameter set used at another point.)
[0022] Once constraints satisfying the similarity criterion are
chosen to remove parameters as discussed above, the third operation
of the determination step is performed using a suitable algorithm
with the constraints as modified by the new constraints to optimize
some or all of the remaining parameters. Any parameter that has not
been removed and is not being optimized over is considered to be
constrained. It is generally beneficial not to constrain parameters
which have not been removed and which have a significant effect on
the performance metric or which are coupled strongly with variables
that are being optimized. This benefit is especially true if an
algorithm is used as part of steps 3 or 4 that is adequate to
optimize such parameters without significant additional computation
time. The parameters optimized in the third operation are
advantageously (but not necessarily) those that have a significant
effect on the signal strength. Such parameters are deemed to have a
significant effect on the signal strength if the dependency of the
performance metric over the allowed values of such parameter varies
more than 10 % at any geographic point in the cell with the cell
defined as for the similarity criterion. Thus, for example, signal
strength associated with a given antenna has a value that generally
depends on a multiplicity of the remaining parameters. If for some
choice of all the parameters but the one under consideration the
signal strength changes more than 10% then such parameter under
consideration has significant effect on such signal strength. This
parameter is considered to have a significant effect on the signal
strength even if there also exists a set of parameters such that
over the full range of the parameter being considered the signal
strength does not change by 10%.
[0023] Typically there are commercially available algorithms for
use in such optimization operations. For example, algorithms such
as the greedy algorithm are useful for optimizing cell site
locations while algorithms such as genetic or simulated annealing
based algorithms are suitable for optimizing antenna orientation
parameters An algorithm is considered acceptable if it improves the
performance of a system.
[0024] In the third operation using a suitable algorithm parameters
being allowed to vary under the new constraints are optimized. The
optimized values (or range of values determined in the third
operation) are employed either 1) to guide the choice of such
parameters in the fourth operation or 2) to provide a further
constraint in the fourth operation. In the former case an optimized
value is obtained and used in the subsequent step or the optimized
value is employed as a guide for choosing a fixed value for such
parameter in the fourth operation. (Such fixed value obtained from
such guidance in the context of this invention is also considered
an optimized value.) In the latter case a plurality, e.g. a range,
of optimized values are obtained for a parameter, and, in the
fourth operation such parameter is constrained to such plurality of
values and optimized together with parameters removed in the second
operation. The parameters removed by imposing the additional
constraint(s) are optimized using appropriate algorithms (not
necessarily the same as used in the third operation) with the
parameters optimized in the third operation being set to values
using the result of the third operation. In essence, in the third
operation, parameters not removed from consideration by the new
constraints are optimized and in the fourth operation these
optimized parameters are fixed (or constrained) while all, or a
portion of, the parameters removed in the third operation are
optimized under the original set of constraints. Clearly the values
obtained in the third operation depend on the variety of things
including the algorithm employed. By use of one or a multiplicity
of algorithms it is possible to determine in the third operation a
range of parameter values that improve the performance metric under
consideration. Use of such optimized parameters in the fourth step
while optimizing all or a portion of the parameters removed in the
second operation allow improved optimization per unit time.
* * * * *