AMR Fleet Behavior Clustering for Deviation Warnings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous Mobile Robots (AMRs) face uncertainties due to inherent limitations in computer vision algorithms and sensors, leading to deviations from expected behaviors, especially when working in fleets, impacting task execution and reliability in logistics, manufacturing, and storage environments.
Innovation Solution
A system using machine-learning models to analyze historical and real-time data from AMRs, generating clusters of expected behavior, resultant vectors, and Resultant of Resultant Vectors (RoRs) to identify deviations, enabling rapid detection and classification of behavioral scenarios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AMRs use computer vision algorithms and sensors to navigate autonomously, then they can avoid obstacles and handle collisions, but inherent uncertainties in these algorithms and sensors lead to deviations from expected behaviors
Solution Approach 1:
The system continuously monitors AMR operational data and compares it against expected behavior patterns, providing feedback when deviations are detected. This enables real-time correction and intervention to maintain reliable task execution despite sensor and algorithm uncertainties.
Solution Approach 2:
The patent introduces a behavioral analysis system as an intermediary layer between the AMR's sensor/algorithm subsystems and the task execution system. This intermediary monitors and interprets operational data to detect deviations before they impact task reliability.
2Productivity
If a fleet of AMRs executes integrated tasks concomitantly, then productivity increases, but coordination complexity increases making it harder to detect and manage behavioral deviations
Solution Approach 1:
The system segments the monitoring of each AMR into independent behavioral clusters, analyzing operational data for each robot separately while maintaining overall fleet coordination. This reduces the complexity of managing deviations in concurrent tasks by breaking down the monitoring problem into manageable segments.
Solution Approach 2:
The behavioral analysis system serves multiple functions simultaneously: it monitors individual AMR behaviors, detects deviations, classifies behavioral scenarios, and provides alerts. This multi-functional approach manages coordination complexity while maintaining high fleet productivity.
3Reliability
If the system monitors all AMRs in real-time to detect behavioral deviations, then reliability improves, but computational resources and time consumption increase
Solution Approach 1:
The system applies partial monitoring by focusing computational resources on detecting specific deviation patterns rather than analyzing every aspect of AMR operation equally. It processes operational data to the extent necessary for reliable deviation detection without excessive computational overhead.
Solution Approach 2:
The system pre-defines behavioral clusters and expected behavior patterns before monitoring begins. This preliminary preparation enables faster real-time deviation detection by comparing operational data against pre-established criteria rather than performing complex analysis during monitoring.
Data Source
AI summary
One example method includes receiving real-time operational data related to the operation of Autonomous Mobile Robots (AMRs) belonging to an AMR group. Clusters of expected behavior, for the AMRs, are accessed. The clusters were generated using historical operational data. Each cluster defines a possible behavioral scenario for each AMR and includes a cluster boundary that defines a limit of the expected behavior and a cluster centroid that defines an average of expected behavior of each AMR. Resultant vectors that extend from the cluster centroid to the most recent operational point of each AMR are generated. A predetermined phase threshold value is used to determine when two or more of the resultant vectors are close to each other. The close resultant vectors are grouped to generate Resultant of Resultant Vectors (RoRs). The RoRs are used to identify behavioral scenarios of the AMRs.