Automatic generation of control decision logic for complex engineered systems from dynamic physical model

Abstract

Possible input value combinations of a prediction of an engineered system are iterated over, comprising, for a possible input value combination: selecting an action to perform on the engineered system for the possible input value combination, comprising: performing a plurality of predictions of the engineered system scored by evaluating an objective function associated with the engineered system and using the possible input value combination and a corresponding plurality of actions. The action is selected from the corresponding plurality of actions, the selection being based at least in part on scores of the plurality of predictions. A rule specifying a corresponding set of one or more rule conditions that is met when the possible input value combination is matched and a corresponding action associated with the rule as a selected action is generated. The generated set of rules to be stored or further processed is output.

1 . A system, comprising:
 a processor configured to:
 iterate over possible input value combinations of a prediction of an engineered system, including to:
 for a possible input value combination:
 select an action to perform on the engineered system for the possible input value combination, at least in part by scoring a prediction; and 
 generate a rule specifying a corresponding set of one or more rule conditions that is met when the possible input value combination is matched and a corresponding action associated with the rule as a selected action; and 
 
 
 output the generated set of rules to be stored or further processed; and 
   a memory coupled to the processor and configured to provide the processor with instructions.   
     
     
         2 . The system of  claim 1 , wherein given the engineered system at an initial state and a prediction action, the prediction substantially determines a later state of the engineered system. 
     
     
         3 . The system of  claim 1 , wherein the engineered system is designed to have predictable performance relative to an application requirement that is captured by an objective function. 
     
     
         4 . The system of  claim 1 , wherein the processor is further configured to select a second action to perform on the engineered system for the possible input value combination in an iteration. 
     
     
         5 . The system of  claim 1 , wherein the action includes setting a set point of a controller. 
     
     
         6 . The system of  claim 1 , wherein the selecting of the action to perform is based at least in part on a prediction and the objective function. 
     
     
         7 . The system of  claim 1 , wherein the objective function indicates whether the engineered system is behaving as desired. 
     
     
         8 . The system of  claim 1 , wherein the objective function quantifies a difference between a simulated behavior and a desired behavior. 
     
     
         9 . The system of  claim 1 , wherein the generated set of rules is further processed, including by using the generated set of rules for subsequent execution of a control system to control the engineered system. 
     
     
         10 . The system of  claim 1 , wherein the input values are discretized into a smaller number of possible values. 
     
     
         11 . The system of  claim 1 , wherein the input values are discretized into a smaller number of possible values and the discretized values are used in iteration over input value combinations. 
     
     
         12 . The system of  claim 1 , wherein to select the action comprises to:
 iterate over potential actions for the possible input value combination, including to:
 predict behavior of the engineered system from an initial state consistent with the possible input value combination to a predicted state at a subsequent time, assuming a potential action under consideration is taken; and 
 evaluate the predicted state according to the objective function; and 
   select the potential action as the action with a preferred result according to the objective function.   
     
     
         13 . The system of  claim 12 , wherein to predict behavior of the engineered system includes to simulate the engineered system from a state consistent with the possible input value combination forward in time. 
     
     
         14 . The system of  claim 1 , wherein a control decision logic is divided into pipeline stages. 
     
     
         15 . The system of  claim 14 , wherein the generated set of rules correspond to one of the pipeline stages. 
     
     
         16 . The system of  claim 14 , wherein the possible input value combination corresponds to one of the pipeline stages. 
     
     
         17 . The system of  claim 14 , wherein the selected action corresponding to a first pipeline stage is included as part of a second possible input value combination corresponding to a second pipeline stage. 
     
     
         18 . The system of  claim 1 , wherein a pruning function is used to eliminate some of the possible input value combinations. 
     
     
         19 . A method, comprising:
 iterating over possible input value combinations of a prediction of an engineered system, including to:
 for a possible input value combination:
 selecting an action to perform on the engineered system for the possible input value combination, at least in part by scoring a prediction; and 
 generating a rule specifying a corresponding set of one or more rule conditions that is met when the possible input value combination is matched and a corresponding action associated with the rule as a selected action; and 
 
   outputting the generated set of rules to be stored or further processed.   
     
     
         20 . A computer program product embodied in a non-transitory computer readable medium and comprising computer instructions for:
 iterating over possible input value combinations of a prediction of an engineered system, including to:   for a possible input value combination:   selecting an action to perform on the engineered system for the possible input value combination, at least in part by scoring a prediction; and   generating a rule specifying a corresponding set of one or more rule conditions that is met when the possible input value combination is matched and a corresponding action associated with the rule as a selected action; and   outputting the generated set of rules to be stored or further processed.