Pervasive model adaptation: the integration of planning and information gathering in dynamic production systems

A computer-based system for pervasive model adaptation, comprising:

• at least one memory including computer executable instructions embodying:
• (i) a model adaptation engine configured to:
• receive observation data produced by at least one sensor of a production system and
• calculate changes to be made to a system model of the production system based on the received observation data from the production system;
• (ii) a system model module adapted to dynamically update the system model based on the calculated changes from the model adaptation engine, and
• a planner module adapted to update production plan based at least partially on the updated system model from the system model module, the production plan including a sequence of actions;
• and,
• at least one processor executing the computer executable instructions;
• wherein the planner module includes:
• a constraint module configured to generate directed graph based on internal constraints of the production system;
• an information criterion module configured to generate heuristics based on the system model;
• and
• a search module configured to update a production plan based on the directed graph from the constraint module and the heuristics from the information criterion module;
• wherein the search module is adapted to:
• perform a best-first search algorithm on the directed graph generated by the constraint module by utilizing an information content function as the evaluation function for the best-first search algorithm to produce a path that maximizes information content;
• and
• set the path produced by the best-first search algorithm as the dated production plan;
• and
• wherein the information content function utilizes a Kalman filter.

A computer-based system for pervasive model adaptation, comprising:

• at least one memory including computer executable instructions embodying:
• (i) a model adaptation engine configured to:
• receive observation data produced by at least one sensor of a production system, and
• calculate changes to be made to a system model of the production system based on the received observation data from the production system;
• a system model module adapted to dynamically update the system model based on the calculated changes from the model adaptation engine;
• and
• a planner module adapted to update a production plan based at least partially on the updated system model from the system model module;
• and
• at least one processor executing the computer executable instructions;
• wherein the production plan updated by the planner module is optimized for both productivity and maximum information content value;
• wherein the planner module includes:
• a constraint module configured to generate a directed graph based on internal constraints of the production system;
• an information criterion module configured to generate heuristics based on the system model;
• and
• a search module configured to update a production plan based on the directed graph from the constraint module and the heuristics from the information criterion module;
• wherein the search module is adapted to:
• perform a best-first search algorithm on the directed graph generated by the constraint module by utilizing an information content function as the evaluation function for the best-first search algorithm to produce a path that maximizes information content;
• and
• set the path produced by the best-first search algorithm as the updated production plan;
• and,
• wherein the information content function utilizes a Kalman filter.