Exoskeleton Intent Recognition Using State-Based Classifier Replacement

Solution Overview

Problem

Conventional intent recognition systems for exoskeletons face challenges in accuracy and delay due to sensor noise and the need for expert supervision, and they fail to adapt effectively to individual user behavior.

Innovation Solution

The development of data-driven intent recognition programs that use sensor data from exoskeleton systems to adapt and refine classification rules automatically, allowing for unsupervised refinement and customization of user intent recognition, reducing the need for human interaction and improving accuracy and responsiveness.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional intent recognition systems use sensor data with expert supervision, then classification rules can be established, but accuracy is reduced due to sensor noise and the system cannot adapt to individual user behavior

Engineering Contradiction:

Improveintent recognition accuracyVSAvoidadaptation to individual user behavior

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system automatically refines classification rules using unsupervised learning algorithms that process sensor data without requiring expert annotation. The exoskeleton system self-adjusts by identifying patterns in user behavior data and updating intent recognition models autonomously, eliminating the need for continuous expert supervision while improving both accuracy and adaptability to individual users

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system implements a feedback loop where classification results are continuously evaluated against actual user behavior outcomes. This feedback mechanism allows the system to detect errors in intent recognition, adjust classification thresholds, and refine rules over time based on accumulated data, thereby improving accuracy while adapting to individual user patterns

Inventive Principle:

Principle #23Feedback

2Adaptability or versatility

If expert supervision is used to establish classification rules, then intent recognition can be implemented, but the system fails to adapt effectively to individual user behavior

Engineering Contradiction:

Improvecustomization to user behaviorVSAvoidunsupervised refinement capability

Core Design Contradiction:

Adaptability or versatilityVSExtent of automation

Solution Approach 1:

The system automatically refines classification rules using unsupervised learning algorithms that process sensor data without requiring expert annotation. The exoskeleton system self-adjusts by identifying patterns in user behavior data and updating intent recognition models autonomously, eliminating the need for continuous expert supervision while improving both accuracy and adaptability to individual users

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The classification rules are designed to be dynamic rather than static, automatically adjusting to changing user behaviors and patterns. The system evolves its recognition models over time by incorporating new data, allowing it to adapt to individual user characteristics and behavioral changes without requiring re-calibration by experts

Inventive Principle:

Principle #15Dynamics

3Productivity

If conventional systems rely on fixed classification rules, then implementation is straightforward, but accuracy and responsiveness are reduced

Engineering Contradiction:

Improveresponsiveness of intent recognitionVSAvoidrecognition accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The classification rules are designed to be dynamic rather than static, automatically adjusting to changing user behaviors and patterns. The system evolves its recognition models over time by incorporating new data, allowing it to adapt to individual user characteristics and behavioral changes without requiring re-calibration by experts

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system continuously processes sensor data and refines classification rules in real-time operation, rather than requiring periodic re-calibration or updates. This continuous learning process ensures that the system maintains high accuracy and responsiveness by constantly adapting to the latest user behavior patterns

Inventive Principle:

Principle #20Continuity of useful action

Data Source

PatentUS11259979B2System and method for user intent recognition

Publication Date: 2022.03.01 ROAM ROBOTICS INC

AI summary

A method of operating an exoskeleton system that includes determining a first state estimate for a current classification program being implemented by the exoskeleton system; determining a second state estimate for a reference classification program; determining that a difference between the first and second state estimate is greater than a classification program replacement threshold; generating an updated classification program; and replacing the current classification program with the updated classification program based at least in part on the determining that the difference between the first and second state estimates is greater than the classification program replacement threshold.