COMPUTATIONAL EFFICIENCY OF NONITERATIVE
PREDICTIVE COORDIATION
K. Stoilova
1
, T. Stoilov
1
1
Institute of computer and communication systems, Bulgarian academy of sciences, 1113 Sofia,
Acad.G.Bontchev str. BL.2,
kstoilova@hsi.iccs.bas.bg
,
todor@hsi.iccs.bas.bg
Abstract
– It is presented a hierarchical
optimization model for solving optimization
problem, which apply analytical relations for the
problem solution. Thus the optimization problem
is solved without iterative and recursive
calculations, speeding up the problem solution.
Such
noniterative
computational
technique
benefits the application of real problem solution,
real time system management and on-line decision
making.
Keywords
– Multilevel hierarchical systems,
optimization, coordination, real-time control
1.
INTRODUCTION
The drawback of the hierarchical approaches, applied
for solving control problems in hierarchical systems,
based on the iterative nature of solving the
optimization problems on different levels. This
iterative manner of computation insists multiple data
transfer of intermediate results between the system
levels [1,2].
For the case of distributed control
systems, these iterative calculations are related with
communication data transfer between the distributed
subsystems, without performing any control actions.
Thus, the iterative solution delays the implementation
of the optimal control decision. This peculiarity of the
hierarchical
system
operation
prevents
the
implementation of the hierarchical system approach
for the on-line system control [4, 5]. To overcome the
delay in hierarchical system management a
noniterative goal control strategy for the hierarchical
systems has been developed [8]. Its peculiarity is
based on the reduced data transfer between the
subsystems and the coordinator in restricted
sequence: prepositions of the local subsystems to the
coordinator and approval or correction by the
coordinator.
The Multilevel hierarchical theory is based on
decomposition approaches for solving the multilevel
optimization problems. The decomposition technique
allows the original complex optimization problem to
be reduced to a set of low order optimization
subproblems, coordinated by the coordination
problem to generate the components of the global
solution of the original problem. Thus, instead of
direct solution of a high ordered and complex
optimization problem, the multilevel theory manages
and coordinates the solutions of low order
optimization subproblems giving solutions to the
initial global optimization problem. The positive
influences of the hierarchical approach address the
large-scale
systems
by
decomposition
and
decentralization of the decision-making and control
processes. Hence, a complex high dimensional
distributed system operates under the control of a set
of low-ordered control units hierarchically connected.