|
 |
|||||||||||
|
|
Real-Time Optimizer IC with Programmable Feasibility Region Team: Rafael Domínguez-Castro and Angel Rodríguez-Vázquez Date: May-1990
Multiparameter optimization is the process
of finding a point of a parameter space x = {x1,
x2, ... xN}T
where a cost function
The fastest way to realize optimization in practice is through the implementation of a companion dynamic gradient system,
Which for given initial point xo obtains the nearest point where the cost function has a local minimum -- the only problem solution for monotonic cost functions.
This companion dynamic system can be implemented as a program on a digital computer, and then solved in serial form -- not suitable for applications where high-speed is a must, as for instance robotics, satellite guidance, etc. This chip parallelizes the solution procedure through the use of a dedicated computing artifacts whose internal operation is analog, while its control is fully digital.
The chip uses switched-capacitor circuit design techniques and encompasses an innovative multiplexing architecture to reduce the circuitry area overhead and thus, increase the complexity of problems that can be mapped on a single piece of silicon. For more information please see [Rodr88], [Rodr90], [Domi92], [Domi93].
References:
Images List:
|
|
|
(x)
is minimized. Most commonly, this cost function contains information about
the problem target and the constraints that the solution point has to
meet (constrained optimization). These constraints define the a region
of the design space where the solution point has to be comprised -- called
feasibility region. 