Auto-Generated Sensor Data Event Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for managing inventory in facilities, such as warehouses and retail stores, face challenges in accurately detecting and verifying events like item picking or returns, especially when confidence levels are low, leading to inefficiencies and potential errors in updating virtual shopping carts and resource allocation.
Innovation Solution
The implementation of a system that uses multiple hypothesis sources, including artificial neural networks and human associates, to process sensor data from cameras, RFID tags, and weight sensors, calculates confidence values, and determines whether to escalate events for verification by additional resources based on likelihood predictions from a trained model, promoting high-confidence results and optimizing resource use.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple hypothesis sources are used to verify events, then accuracy of event detection is improved, but device complexity and resource consumption increase
Solution Approach 1:
A trained machine learning model serves as an intermediary between sensor data collection and full verification processing. The model predicts likelihood of agreement between hypothesis sources, allowing the system to selectively engage additional verification only when necessary, thus reducing overall system complexity while maintaining high accuracy
Solution Approach 2:
The system performs preliminary analysis using a trained model to predict whether additional hypothesis sources would agree with initial detections. This preliminary action prevents unnecessary engagement of additional verification resources, simplifying the system architecture while preserving detection accuracy for uncertain cases
2Measurement precision
If additional hypothesis sources are engaged for verification, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The system applies partial verification by engaging additional hypothesis sources only when the trained model predicts low confidence or disagreement. For high-confidence detections, the system accepts partial verification from the first hypothesis source alone, significantly reducing verification time while maintaining precision for uncertain cases
Solution Approach 2:
The trained model performs preliminary assessment of confidence levels before engaging additional verification resources. This preliminary action enables the system to quickly resolve most events without time-consuming full verification, while reserving additional hypothesis sources only for borderline cases
3Reliability
If human associates are used to verify low-confidence events, then reliability is improved, but productivity decreases
Solution Approach 1:
Human associates are engaged only for partial verification of low-confidence events predicted by the trained model. The majority of high-confidence events are processed automatically without human intervention, maintaining high productivity while ensuring reliability for uncertain cases through selective human verification
Solution Approach 2:
The trained machine learning model acts as an intermediary that filters events requiring human verification. By predicting which events are likely to be disagreed upon, the model directs only necessary cases to human associates, preserving their productivity while maintaining overall system reliability
4Measurement precision
If confidence values are calculated and thresholds applied, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system changes the parameter of confidence assessment from binary (verified/not verified) to a continuous confidence value scale. By calculating and comparing confidence values against thresholds, the system achieves precise measurement of detection reliability while using simple threshold comparison logic to avoid excessive processing complexity
Data Source
AI summary
This disclosure describes systems and techniques for detecting events, determining a result of each respective event using a first hypothesis source, and calculating a likelihood that a second (and/or additional) hypothesis source would determine the same result of the respective event. The calculated likelihood may then be used to be determine whether to request that the second hypothesis source determine the result of the event, determine an amount of resources of the second hypothesis source to use to make this determination, and/or like.