# US 2012/0254081 A1 · Optimization control system

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2012/0254081 A1
- Document ID: US20120254081A1
- Status: Published application
- Kind: A1
- Application number: 13/421,206
- Filed: 2012-03-15
- Published: 2012-10-04
- Earliest claimed: 2011-03-30
- Assignees: Honda Motor Co., Ltd.
- Inventors: Takumi Kamioka; Masanori Takeda; Mitsuhide Kuroda; Shigeru Kanzaki
- Web page: https://patentgrep.com/patent/US20120254081A1

## CPC

- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G05B 13/024: {in which a parameter or coefficient is automatically adjusted to optimise the performance}
- G06F 17/11: for solving equations {, e.g. nonlinear equations, general mathematical optimization problems (optimization specially adapted for a specific administrative, business or logistic context G06Q10/04)}
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Provided is an optimization control system in an attempt to improve searching accuracy of an optimal solution defining a behavior mode for a control subject. A plan storing element 120 is configured to obtain a current update result of a joint probability distribution p(u, x) on the basis of a current update result of a probability distribution p(x) from a state estimating element 110 and a current update result of the conditional probability distribution p(u|x) from a behavior searching element 200. The behavior searching element 200 is configured to determine the conditional probability distribution p(u|x) as a current basis for obtaining the current update result of the conditional probability distribution p(u|x) on the basis of the current update result of the probability distribution p(x) from the state estimating element 110 and a previous update result of the joint probability distribution p(u, x) from the plan storing element 120.

## Claims

### Claim 1

An optimization control system configured to determine an optimal behavior for a control subject, comprising:
a state estimating element configured to update a probability distribution p(x) of a state x according to a determination result of the state x of the control subject each time;
a plan storing element configured to update a joint probability distribution p(u, x) of the state x and a behavior u according to the determination result of the state x of the control subject each time; and
a behavior searching element configured to iterate a generation of multiple behavior candidates ui in accordance with a conditional probability distribution p(u|x), an evaluation of a goodness of fit fi in consideration of an evaluation function f(u|x) of each of the multiple behavior candidates, a selection of a behavior candidate in accordance with the evaluation result, and an update of the conditional probability distribution p(u|x) so as to approach a distribution characteristic of a selected behavior candidate to a shape characteristic of the evaluation function f(u|x) so as to obtain a current update result of the conditional probability distribution p(u|x) and to determine an optimal behavior u* as a behavior which the control subject should take in consideration of the goodness of fit,
wherein
the plan storing element is configured to obtain a current update result of the joint probability distribution p(u, x) on the basis of a current update result of the probability distribution p(x) and the current update result of the conditional probability distribution p(u|x) by the behavior searching element, and
the behavior searching element is configured to determine the conditional probability distribution p(u|x) as a current basis for obtaining the current update result of the conditional probability distribution p(u|x) on the basis of the current update result of the probability distribution p(x) from the state estimating element and a previous update result of the joint probability distribution p(u, x) from the plan storing element.

### Claim 2 (depends on 1)

The optimization control system according to claim 1, wherein
the state estimating element is configured to update the probability distribution p(x) defined as a linear combination of multiple unimodal probability distributions, and
the behavior searching element is configured to update the conditional probability distribution p(u|x) defined as a linear combination of multiple conditional probability distributions having different dependence of mean value with respect to the state x.

### Claim 3 (depends on 1)

The optimization control system according to claim 1, wherein
the behavior searching element is configured to update the conditional probability distribution p(u|x) so as to maximize an expected value E which is an integral value of a product of the evaluation function f(u|x) and the conditional probability distribution p(u|x).
