Amazon Technologies, Inc.

Alexander Clark Prater

Dilip Kumar

Sridhar Boyapati

Mackenzie Smith

Suzanne C Reynolds

Kevin Alexander Lee

Chang Yuan

Xiaofeng Ren

Elie Micah Kornfield

Korwin Jon Smith

David William Bettis

Sudarshan Narasimha Raghavan

Alexander Michael McNamara

Eric C Young

Aaron Craig Thompson

Brett Allen Skaloud

Udit Madan

Navid Shiee

David Echevarria Ignacio

Félix Joseph Étienne Pageau

Christopher Andrew Stephens

Ismael Medrano

Yung-Chun Lin

Gary Paolo Raden

Casey Louis Thurston

Nipun Dureja

Kenneth King-Fung Lee

Felix F Anthony

Ivan Avramov

Samir Kumar

Matthew Rice Carlson

James E Bacus

Priya Padmanabhan

Prajakta Vaidya

Victor Gregory Estevez

Cary R Champlin

Anthony B Williams

Dean W Webster

Benjamin Lloyd Shapiro

Russell Dicker

Ameet Nirmal Vaswani

Natalie Thuy-Tien Nguyen

Andrew V Harbick

Ming-Hsiu Wang

Joseph Papa

Gregory Martin

Matt Round

John Benjamin Compton

John B Compton

This disclosure describes a system for tracking removal or placement of items at inventory locations with a materials handling facility. In some instances, a user may remove an item from an inventory location and the inventory management system may detect that removal and update a user item list associated with the user to include an item identifier representative of the removed item. Likewise, if the user places an item at an inventory location, the inventory management system may detect that placement and update the user item list to remove an item identifier representative of the placed item.

Detecting inventory changes

A system may use sensor data from a facility to generate tentative values associated with an event, such as the identification of an item removed from a shelf of the facility. A confidence value associated with each of the tentative values may be less than a confidence threshold. In response, inquiry data seeking confirmation of a tentative value from an associate is generated and sent to one or more associates in the facility. Responses from the associates are collected to determine a selection of one of the tentative values. The selected tentative value is designated as output data for the system. Thereafter, the output data and the original sensor data are designated as training data, which can then be used to train or update machine learning systems. Subsequent use of the updated machine learning systems can yield more accurate results.

Updating machine learning systems using previous output data

This disclosure describes techniques for utilizing sensor data to determine the results of events within a system. Upon receiving sensor data indicative of an event, the sensor data is analyzed to determine a result of the event. For instance, a user may identified upon entry to a facility, may pick items from shelves in the facility, and may be identified upon exit. Contents of a virtual shopping cart of the user may be maintained based on analysis of the sensor data as the user picks the items. When a user exits, a payment instrument of the user may be charged for a cost of the items and a receipt may be sent to a device of the user. A state-invariant time, representing a time before which all events have been processed, may be maintained such that no shopping sessions are finalized prior to the state-invariant time to avoid mistaken charges.

Maintaining accurate cart-state using auto-generated sensor data

A bulk data distribution system in which, instead of multiple data consumers contending to access the same data sets from a primary data store, one or more producers capture snapshots (states of the data sets at particular points in time) and upload the snapshots to an intermediate data store for access by snapshot consumers. The snapshot consumers may download the snapshots to generate local versions of the data sets for access by one or more data processing applications or processes. A snapshot producer may periodically generate full snapshots of a data set, and may generate one or more incremental snapshots of the data set between full snapshots. A snapshot consumer may bootstrap a local data set from a full snapshot and one or more incrementals, and may maintain state of the local data set by accessing new snapshots uploaded by the producer.

Bulk data distribution system

This disclosure describes techniques for utilizing sensor data to automatically determine the results of events within a system. Upon receiving sensor data indicative of an event, the techniques may analyze the sensor data to determine a result of the event, such as that a particular user associated with a user identifier selected a particular item associated with an item identifier. The techniques described herein may determine when sensor data is to be stored for a default amount of time, and when sensor data is to be stored for a greater amount of time. By retaining data in an intelligent manner, the system is able to function accurately without unduly taxing storage constraints.

Automated sensor data retention

Described is an interactive data processing system configured to facilitate selection by a human associate of tentative values generated by an automated system from sensor data. In one implementation, an event may take place in a materials handling facility. The event may comprise a pick or place of an item from an inventory location, movement of a user, and so forth. The sensor data associated with the event is processed by an automated system to determine tentative values associated with the event. In some situations, an uncertainty may exist as to which of the tentative values accurately reflects the actual event. The sensor data and tentative values may be provided to a human associate. The associate may select one of the tentative values. The selected tentative value and the sensor data may be used to train the automated system.

Interactive data processing system

An inventory location, such as a shelf, may be used to stow different items. The shelf is divided into a plurality of adjacent lanes, some of which may be grouped into a cluster. Items are placed upon the lanes of the cluster. Each lane includes one or more weight sensors to obtain weight data about a load on the lane. Based on the weight data and physical layout data for the shelf, one or more hypotheses may be determined for describing interactions, such as a pick or place of an item, occurring at the lanes of a cluster. One of the hypotheses may be selected as a solution for describing the interaction and used to maintain inventory quantities of items at the shelf.

Inventory management using weight sensors and physical layout data

Detecting item interaction and movement

An inventory location such as a shelf may be used to stow different types of items. The shelf may be equipped with a plurality of lanes arranged parallel to one another. Items may rest upon two or more of the lanes. Each lane includes a weight sensor that provides weight data about a load on the lane. Based on the weight data and item data indicative of individual weights of those items stowed on the shelf, interaction data indicative of an activity such as a pick or place of an item may be determined. The interaction data may specify what item was picked or placed, quantity of that item that was picked or placed, and so forth. Data from other sensors, such as a camera, may be used to determine the interaction data.

Item management system using multiple scales

Various embodiments of a system and method for authoritative item identification based on facility operations are described. Embodiments may include an authoritative item identification component configured to, for each particular item of multiple items in inventory of a materials handling facility, determine multiple identifiers for the particular item, where one or more of the multiple identifiers are electronically obtained from a unit of the particular item during operations of the materials handling facility. The authoritative item identification component may, for each of the multiple identifiers, determine a quantity of instances in which that identifier was electronically obtained from a given unit of the particular item. The authoritative item identification component may, based on a statistical analysis of the quantities, determine one of the multiple identifiers as being an authoritative identifier for the particular item. The authoritative item identification component may identify one or more units based on the authoritative identifiers.

System and method for authoritative item identification based on facility operations

A package handling system for use with a plurality of packages having package identification thereon. The system of one embodiment comprises processing stations where packages having heterogeneous contents are processed. One or more identifiers are positioned at the processing stations to detect the package identifications associated with the packages. A controller is coupled to the identifier and has access to first and second package data. The first package data related to an actual condition of the package, and the second package data related to acceptable package conditions. A labeler coupled to the controller is configured to apply a first label to the package if the first and second package data indicate that the package is in an acceptable condition or a second label if the first and second package data indicate that the package is in a condition requiring further processing.

Package handling system with kickout labeling

Growing community of inventors

Seattle, WA, United States of America

Alexander Clark Prater