Granted Patent B2

US 7,979,293 · App. 11/238,825 · Granted Jul 12, 2011

Abstract

Claims (71)

1. A method for managing operations of a production system, comprising:

partitioning the system into a plurality of subsystems;

for each subsystem, at least one computing device, coupled to a memory executing the steps of:

creating an agent to associate with the subsystem;

providing to the agent one or more goals for the subsystem; and

providing to the agent an overall objective function for the system; and

executing, using the at least one computing device, an optimization algorithm involving the agents to manage system operations based on the goals and the overall objective function, wherein the optimization algorithm implements a sampled fictitious play algorithm that includes:

providing a policy to the agents;

logging the policies of the agents to maintain a historical record of policies;

and

in successive iterations:

generating a plurality of joint policies of the agents from the historical record;

the agents independently determining policies assuming other agents jointly choose one of the plurality of joint policies at random;

synthesizing the policies determined by the agents into a synthetic joint policy;

adding to the historical record by logging the policies determined by the agents; and

evaluating the expected value of the synthetic joint policy based on the goals and the overall objective function;

wherein iterations continue until a stopping criterion is met.

2. A method as described in claim 1 , wherein providing to the agent an overall objective function for the system comprises:

providing to the agent a maximization of production throughput for the system.

3. A method as described in claim 1 , further comprising:

providing a random agent for a subsystem;

wherein the random agent affects operation of the subsystem according to a probability distribution.

4. A method as described in claim 1 , wherein the stopping criterion comprises:

reaching a predetermined iteration limit.

5. A method as described in claim 1 , wherein the sampled fictitious play algorithm further includes:

retrieving current system information.

6. A method as described in claim 1 , wherein the stopping criterion comprises:

achieving a predetermined number for the expected value of the synthetic joint policy.

7. A method as described in claim 5 , further comprising:

recording a joint policy created during use of the sampled fictitious play algorithm to produce a recorded joint policy having a corresponding individual policy for the agents;

wherein providing a policy to the agents comprises:

providing the corresponding policy of a recorded joint policy to the agents.

8. A method for optimizing the timing and duration of speed reductions of a plurality of subsystems of a production system, comprising performing, by at least one computing device, coupled to a memory, the steps of:

providing a set of agents associated with the subsystems, one agent corresponding to one subsystem;

providing a throughput target for the agents; and

providing a common production throughput objective for the agents;

wherein an agent determine policies and participates in a sampled fictitious play algorithm to manage system speed reductions based on the throughput targets and the common production throughput objective, the sampled fictitious play algorithm comprises:

providing a policy to the agents;

logging policies of the agents to maintain a historical record of policies; and

in successive iterations:

generating a plurality of joint policies of the agents from the historical record;

the agents independently determining a policy assuming other agents jointly choose one of the plurality of joint policies at random;

synthesizing the policies determined by the agents into a synthetic joint policy;

adding to the historical record by logging the policies determined by the agents; and

evaluating the expected value of the synthetic joint policy based on the throughput targets and the common production throughput objective;

wherein iterations continue until a stopping criterion is met.

9. A method as described in claim 8 , wherein the agents are configured to manage system operations by making decisions as to the number, time of occurrence and duration of stopping times for its corresponding subsystem.

10. A method as described in claim 8 , wherein the sampled fictitious play algorithm further includes:

retrieving current production system information.

11. A system for managing the timing and duration of speed reductions of a plurality of subsystems of a production system, the system including at least one computing device coupled to a memory having stored thereon a plurality of modules;

an agent module for providing a set of agents associated with the subsystems, one agent corresponding to one subsystem;

a throughput target module for providing a throughput target for an agent;

an objective module for providing, for the agents, a common production throughput objective; and

a sampled fictitious play module for providing instructions of a sampled fictitious play algorithm wherein the agents provided by the agent module participate in the sampled fictitious play algorithm to manage system speed reductions based on the throughput targets provided by the throughput target module and the common production throughput objective provided by the objective module, and wherein the sampled fictitious play module includes:

an initialization module for providing a policy to the agents

a logging module for logging the policies of the agents to create a historical record of policies

a generating module for generating a plurality of joint policies of the agents from the historical record

a determining module for the agents for independently determining a policy assuming other agents jointly choose one of the plurality of joint policies at random;

a synthesizing module for synthesizing the policies determined by the agents into a synthetic joint policy; and

an evaluation module for evaluating the expected value of the synthetic joint policy based on the throughput targets and the common production throughput objective;

wherein the logging, generating, determining, synthesizing, and evaluation modules are executed iteratively until a stopping criterion is met.

12. A system as described in claim 11 , further comprising:

a production interface module for monitoring a status of the production system; and

a control interface module for controlling a subsystem of the production system;

wherein system speed reductions are managed by monitoring the status of the production system, and by controlling the subsystem of the production system.

13. A system as described in claim 11 , wherein the production system has current production system information and the sampled fictitious play module further comprises:

a retrieval module for retrieving current production system information;

wherein the evaluation module is further for evaluating the expected value of the synthetic joint policy based on the throughput targets, the common production throughput objective, and the current production system information.

14. A system as described in claim 13 , further comprising:

a database storage and retrieval module for storing and retrieving historical records of policies previously determined by agents;

wherein the initialization module is further for providing a policy to the agents by retrieving a historical record of a policy previously determined by the agents.

Assignments (12)