Robot Actuator Control via Contact Force Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for controlling robot actuators fail to effectively distinguish between desirable and undesirable environmental contacts, particularly during robot-human interactions and collisions with surrounding objects or people, leading to inadequate open-loop and closed-loop control.
Innovation Solution
A method and device that utilize a sensor system to detect and classify external contact forces and moments using a feature vector analysis, allowing for precise differentiation between desirable and undesirable contact events by assessing time-dependent variables and employing support vector machines or other classification techniques for open-loop and closed-loop control adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If robot actuators are controlled without distinguishing contact types, then control simplicity is maintained, but safety and damage prevention are inadequate
Solution Approach 1:
The patent segments contact events into distinct categories (desirable vs. undesirable contacts) based on sensor data analysis. By dividing the control problem into classification segments, the system can apply different control strategies for different contact types, improving safety without requiring complete redesign of the control architecture.
Solution Approach 2:
The patent introduces an intermediary classification layer between sensor detection and actuator control. This intermediary system analyzes contact characteristics and determines contact type, serving as a mediator that translates raw sensor data into meaningful control decisions, thereby enhancing safety while maintaining manageable control complexity.
2Measurement precision
If sensor system detects all external forces and moments, then measurement completeness is improved, but processing complexity and time increase
Solution Approach 1:
The patent extracts only the essential features from complete sensor data that are necessary for contact classification. By taking out and focusing on critical parameters rather than processing all available sensor information, the system achieves accurate contact detection while reducing processing time and computational burden.
Solution Approach 2:
The patent applies partial action by detecting and analyzing only the minimum necessary sensor parameters required for contact classification. Rather than fully processing all external forces and moments in detail, the system identifies sufficient statistics for classification, achieving measurement precision with reduced processing overhead.
Data Source
AI summary
A method and device for the control and regulation of actuators of a robot, taking environmental contacts into consideration, wherein the robot comprises at least two parts, which are connected by an articulated joint drivable by an actuator. The method comprises: by way of a sensor system, ascertaining and storing a time-dependent variable, as a function of the time, of one or more external contact forces and/or of one or more external moments on the parts, providing a condition for the variable, classifying the feature vector based on predefined categories, which each indicate a contact type between one of the parts or the articulated joint and an object in a surrounding environment, which are each imparted by corresponding external contact forces and/or external contact moments, to generate a classification result, and open-loop and/or closed-loop control of the actuator as a function of the classification result.