AI Action Control Using Fuzzy Weighting for Smooth Decisions

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing artificial intelligence control methods using discrete decisions struggle to ensure smooth control over AI devices, leading to poor action smoothness.

Innovation Solution

The method involves obtaining N-dimensional states of an AI device, using active fuzzy subsets to generate continuous decisions through weighted summation based on membership degrees, ensuring smooth control by converting discrete decisions into continuous outputs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If discrete decisions are used to control AI device actions, then the control model can output clear decisions, but the action smoothness deteriorates

Engineering Contradiction:
Improvedecision clarityVSAvoidaction smoothness
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The patent transforms the discrete decision output into a continuous decision through parameter changes. Specifically, it uses fuzzy subset membership degrees (continuous parameters ranging from 0 to 1) to weight and combine multiple discrete decisions, thereby converting the discrete output into a continuous control signal that ensures smooth AI device actions while maintaining decision clarity.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces fuzzy subset membership degrees as an intermediary between the discrete decisions and the final control output. These membership degrees serve as weighting factors that smoothly blend multiple discrete decisions based on their relevance to the current state, acting as a mediator that transitions from discrete to continuous control without losing decision information.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Ease of operation

If continuous decisions are generated through weighted summation, then action smoothness improves, but the complexity of the control process increases

Engineering Contradiction:
Improveaction smoothnessVSAvoidcontrol process complexity
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The patent segments the control process into distinct modular steps: obtaining states, determining active fuzzy subsets, calculating membership degrees, obtaining discrete decisions from the control model, performing weighted summation, and generating continuous decisions. This segmentation organizes the complex control process into manageable, independent modules that can be implemented and debugged systematically, reducing overall system complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary actions by pre-defining fuzzy subsets and their membership functions before the actual control execution. The active fuzzy subsets and their corresponding membership degrees are determined in advance based on the current state, allowing the weighted summation to be performed efficiently without real-time complex calculations, thus reducing control process complexity.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3719603B1Action control method and apparatus
Publication Date: 2023.04.26 HUAWEI TECH CO LTD
  • EP3719603B1 patent drawingFigure 1~2
  • EP3719603B1 patent drawingFigure 3
  • EP3719603B1 patent drawingFigure 4A

AI summary

The present disclosure provides an action control method and apparatus, and relates to the field of artificial intelligence. The method includes: obtaining states of N dimensions of an artificial intelligence device; obtaining a plurality of discrete decisions based on an active fuzzy subset and a control model that are of a state of each of the N dimensions, where an active fuzzy subset of a state is a fuzzy subset whose membership degree of the state is not zero, each fuzzy subset includes a state interval that corresponds to a same discrete decision in a dimension, the membership degree is used to indicate a degree that the state belongs to the fuzzy subset, and the control model is used to output a corresponding discrete decision based on an input state; performing, based on a membership degree between a state and an active fuzzy subset that are of each dimension, weighted summation on the plurality of discrete decisions, to obtain a continuous decision; and controlling, based on the continuous decision, the artificial intelligence device to execute a corresponding action. An output decision in the present disclosure is a continuous quantity, thereby ensuring smooth control over the artificial intelligence device, and ensuring smoothness of an action.