Safe Control Input Selection Using Mutual Information

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for controlling computer-controlled systems, such as robots and manufacturing systems, face challenges in efficiently learning about the safety of control inputs without violating constraints, particularly in real-world environments with noisy measurements and unknown constraint quantities, and often require excessive interaction with the environment.

Innovation Solution

The proposed method uses Information-Theoretic Safe Exploration (ISE) to determine control inputs based on mutual information between the constraint quantity and other control inputs, optimizing for information gain about safety without relying on discrete domains or hyperparameters like the Lipschitz constant, allowing for efficient exploration of continuous-valued control inputs and improved data efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional safe exploration methods (SafeOpt) are used with Lipschitz continuity assumptions, then safety constraints are satisfied, but data efficiency deteriorates due to excessive interaction with the environment

Engineering Contradiction:
Improvesafety constraint satisfactionVSAvoiddata efficiency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces the mechanical/Lipschitz-based exploration framework with an information-theoretic framework using mutual information. Instead of relying on Lipschitz constants and discrete domain assumptions, the method uses MI to directly measure and optimize the information gained about safety constraints, enabling more efficient exploration with fewer environmental interactions.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the fundamental parameter for exploration from Lipschitz constants to mutual information. By optimizing control inputs based on MI between constraint quantities and control inputs, the system adapts its exploration strategy dynamically, focusing on regions that provide maximum information about safety boundaries rather than following fixed Lipschitz-based trajectories.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If discrete domain assumptions are made for safe exploration, then safety can be guaranteed, but applicability to continuous-valued control inputs deteriorates

Engineering Contradiction:
Improvesafety guaranteeVSAvoidcontinuous domain applicability
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent substitutes the discrete domain framework with a continuous domain information-theoretic framework. By using mutual information, which is defined for continuous random variables, the method naturally extends safe exploration to continuous-valued control inputs without requiring discretization, thereby maintaining both safety guarantees and continuous domain applicability.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Ease of operation

If multiple hyperparameters (e.g., Lipschitz constant) are used for safe exploration, then exploration can be controlled, but system complexity increases

Engineering Contradiction:
Improveexploration controlVSAvoidhyperparameter management
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The patent extracts and removes the Lipschitz constant hyperparameter from the exploration framework. By replacing it with mutual information, the method eliminates the need to manually tune Lipschitz constants while maintaining exploration control. The MI metric inherently adapts to the problem structure without requiring external hyperparameter specification.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4307055A1Constrained controlling of a computer-controlled system
Publication Date: 2024.01.17 ROBERT BOSCH GMBH
  • EP4307055A1 patent drawingFigure 1~2
  • EP4307055A1 patent drawingFigure 3
  • EP4307055A1 patent drawingFigure 4

AI summary

The invention relates to a computer-implemented control method (700) of constrained controlling of a computer-controlled system. The system is controlled according to a control input, which is safe if a constraint quantity resulting from the controlling of the computer-controlled system exceeds a constraint threshold. A current control input is determined based on previous control inputs and corresponding previous noisy measurements. The computer-controlled system is controlled according to the current control input, thereby obtaining a current noisy measurement of the resulting constraint quantity. The current control input is determined based on a mutual information between a first random variable representing the constraint quantity resulting from the current control input and a second random variable indicating whether a further control input is safe.