Data processing: software development – installation – and managem – Software program development tool – Translation of code
Reexamination Certificate
2000-06-14
2001-09-11
Powell, Mark (Department: 2122)
Data processing: software development, installation, and managem
Software program development tool
Translation of code
C717S152000, C703S006000
Reexamination Certificate
active
06289508
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to the field of signal processing and control systems. More particularly, the present invention relates to the dynamic optimization of time varying trajectories to make signal processing and control systems achieve desired outcomes.
2. Background Information
Signal processing and the control of physical systems generally involves obtaining measurements from a physical system in the form of electrical signals and processing the signals in order to bring about a desired result. For example, the control of a physical system typically involves obtaining measurements from the physical system, comparing the measurements with a predetermined control recipe, and adjusting the system inputs, all in real time, in response to the comparison to minimize variations between the measured values and recipe values. During signal processing and control, the signals to be processed or the variables to be controlled are not always directly available for observation and must be inferred from indirect and noisy measurements. The indirect measurements are generally obtained from embedded sensors which contain multiple pieces of information that are dynamically confounded. Extraction of the information of interest requires the use of complex and time consuming calibration procedures and the use of estimating techniques that result in high computation costs. Equipment setup costs are also high since diagnostic measurements must be taken to correlate measured signals to indirect measurements for each piece of equipment.
In some physical systems there are variables which cannot be measured during operation but which affect the ultimate outcome of the system. Thus, some unmonitorable variable (or variables) affect the final system output.
In addition, the measurements that are available from a physical system are not always obtainable at a single time interval (time scale). For example, there may be a first measurement that is obtainable only at a first time scale, a second measurement that is only obtainable at a second time scale, a third measurement that is only obtainable at a third time scale, and so on.
Further, the ultimate variable (or variables) of interest are often only available at a coarse time scale (i.e. a slow rate, for example after each run is completed). In a process for depositing a film on a semiconductor wafer, the thickness of the film deposited cannot be directly measured until the run (or process) is finished.
Moreover, some continuous processes also have variables that must be measured at two or more different time scales. Such variables are sometimes only available at a course time scale and these variables are often the variables that need to be controlled. In a process for controlling the peak power demand in a captive electric generator connected to a power distribution grid, the peak power demand value cannot be directly measured until a set time interval has elapsed.
Often, the task of controlling a system involves not only the control of a single physical system, but the control of a family of similar but not identical physical systems. This situation is most prevalent in high volume manufacturing applications. The characteristics of a single physical system tend to change over time due to equipment degradation and other causes. Moreover, the characteristics of members of a family of physical systems tend to differ from one physical system to another due to equipment manufacturing variations. It is important to account for these differences so that the signal processing or control system may be updated accordingly. Otherwise, the accuracy of the signal processing or control system is compromised.
Current signal processing and control systems do not provide for the rapid calibration of such systems, nor do they provide for the rapid computation of time varying control trajectories to optimize the performance of these systems.
There are a number of related control methods that are collectively known as Model Predictive Control (MPC). MPC computes an optimal trajectory over a finite, but usually long, time interval. The first part (in time) of the optimal trajectory is applied to a system for an interval much shorter than the full time interval and the remaining portion of the optimal trajectory is discarded. In MPC the optimal trajectory applied during the short time period to the system is held constant. At the end of that short time interval, the process is repeated for a second part (in time), i.e., where the first part left off. This process is repeated indefinitely, or as many times as is desired in a continuous process. MPC methods provide continuing control at a single time scale (determined by the short time interval).
Thus, what is needed is an accurate and cost efficient method and apparatus for processing signals generated within a physical system, or family of physical systems, that allow the modification and control of time varying trajectories to optimize the performance of the physical system at different time scales in the face of the issues described above.
SUMMARY OF THE INVENTION
The present invention is a method for modifying and controlling time varying trajectories for signal processing in a physical system, or a family of physical systems. In one embodiment of the present invention the time varying trajectories of a device under control and the related controller are parameterized. For example, the reference (or command) signals and the actuator commands (or control signals) are all parameterized. Next, a vector valued function is computed such that if the reference (or command) signals are acted on by the vector valued function, the outcome parameters result. An estimator that captures the dynamic behavior of the physical system is then built. A cost function that evaluates the output of the physical system is then built. The reference signals are set to a nominal reference trajectory. The physical system is then run using the most recent reference signals. The estimator is then run using valuations for the most recent measured outcome parameters and the most recent control signals to produce an estimated output. The estimated output is then evaluated. If the estimated output is approximately the desired outcome then the system stops. However, if the estimated output is not approximately the desired outcome, then the process continues. The upper and lower bounds for the actuator commands are determined such that the control signals will not reach saturation within these bounds. A constrained nonlinear program is then run to produce new reference signals. The new reference signals are then compared to the previous reference signals. If the new reference signals are approximately the same as the previous reference signals then the system stops. However, if the new reference signals are not approximately the same as the previous reference signals, then the process steps are repeated by running the new reference signals on the physical system to produce a new outcome parameter.
Additional features and benefits of the present invention will become apparent from the detailed description, figures, and claims set forth below.
REFERENCES:
patent: 4599692 (1986-07-01), Tan et al.
patent: 4620286 (1986-10-01), Smith et al.
patent: 5268835 (1993-12-01), Miyagaki et al.
patent: 5424963 (1995-06-01), Turner et al.
patent: 5517594 (1996-05-01), Shah et al.
patent: 5572627 (1996-11-01), Brown
patent: 5579440 (1996-11-01), Brown
patent: 5740033 (1998-04-01), Wassick et al.
patent: 5850356 (1998-12-01), Yamada et al.
patent: 5909381 (1999-01-01), Shome et al.
patent: 6041141 (2000-03-01), Yamamoto et al.
patent: 6041172 (2000-12-01), Shah et al.
patent: 6071317 (2000-06-01), Nagel
patent: 6096085 (2000-08-01), Sammelman
patent: 6167360 (2000-12-01), Erickson et al.
Ganguly et al, “Query optimization for parallel execution”, ACM SGIMOD pp. 9-18, Apr. 1992.*
Ozcan et al, “Dynamic query optimization on a distributed object mamagemnt platform”, ACM CIKM pp. 117-124, Aug. 1996.*
Visweswa
Erickson Mark A.
Gudmundsson Thorkell
Pandey Pradeep
Shah Sunil C.
Blakely & Sokoloff, Taylor & Zafman
Khatri Anil
Powell Mark
Voyan Technology
LandOfFree
Method and apparatus for dynamic optimization does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for dynamic optimization, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for dynamic optimization will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2524472