Inferred determination of interaction using data from weight sensors

Patent metadata

Field	Value
Publication number	US-11263583-B1
Application number	US-201815939061-A
Country	US
Kind code	B1
Filing date	Mar 28, 2018
Priority date	Mar 28, 2018
Publication date	Mar 1, 2022
Grant date	Mar 1, 2022

Abstract

Load cells measure the weight of items on a shelf. Weight changes occur as items are picked from or placed to the fixture. Information about these weight changes is used to determine an estimated location on the shelf of a weight change. Hypotheses are generated using information about where particular types of items are stowed, the weights of those particular types of items, information about the weight changes, and the estimated locations of the weight changes. A model is used to produce confidence values in the hypotheses based on a change in weight measured at a first side and a change in weight measured at a second side of the shelf. A hypothesis with a confidence value that exceeds the threshold may be selected and used to determine interaction data indicative of a quantity picked or placed, type of item, and location on the shelf.

First claim

What is claimed is:

1. A method comprising:
accessing first weight data obtained by a first weight sensor located at a first corner of a shelf;
determining a first weight change at the first corner of the shelf based on a first weight measurement at a first time by the first weight sensor and a second weight measurement at a second time by the first weight sensor;
accessing second weight data obtained by a second weight sensor located at a second corner of the shelf;
determining a second weight change at the second corner of the shelf based on a third weight measurement at the first time by the second weight sensor and a fourth weight measurement at the second time by the second weight sensor;
accessing physical layout data indicative of:
a physical configuration with respect to the shelf of the first weight sensor and the second weight sensor; and
an area of the shelf that is associated with stowage of a type of item;

accessing item data indicative of the type of item designated as being stowed in the area of the shelf;
determining a first set of event hypotheses based on the physical layout data, the item data, the first weight change, and the second weight change, wherein each of the event hypotheses in the first set of event hypotheses is indicative of one or more of:
a possible pick from the shelf of a possible quantity of the type of item from the area of the shelf, or
a possible place to the shelf of a possible quantity of the type of item to the area of the shelf;

determining, using the first weight change and the second weight change as inputs to a model, first confidence values for the event hypotheses in the first set of event hypotheses, wherein the model comprises, for one or more of the event hypotheses in the first set of event hypotheses:
a first probability distribution function based on weight values associated with the type of item as measured at the first corner of the shelf and based on a location of the type of item as stowed on the shelf, wherein the first probability distribution function accepts as input the first weight change and produces as an output a second confidence value indicative of a likelihood that the first weight change is associated with the type of item; and
a second probability distribution function based on weight values associated with the type of item as measured at the second corner of the shelf and based on the location of the type of item as stowed on the shelf, wherein the second probability distribution function accepts as input the second weight change and produces as an output a third confidence value indicative of a likelihood that the second weight change is associated with the type of item;

selecting one or more event hypotheses in the first set of event hypotheses based at least in part on the first confidence values; and
generating, based on the one or more event hypotheses that are selected, data indicative of one or more of a type of an interaction with the shelf, the type of item involved in the interaction, or a quantity of the type of item involved in the interaction.

2. The method of claim 1 , wherein:
the first confidence value for an event hypothesis is based on the second confidence value and the third confidence value.

3. The method of claim 1 , further comprising:
determining, based on the first weight change and the second weight change, an estimated location of a weight change with respect to the shelf;
determining, for the each event hypothesis in the first set of event hypotheses, a fourth confidence value indicative of a correspondence between the estimated location and an area indicated by the each event hypothesis; and
wherein the one or more event hypotheses in the first set of event hypotheses are further selected based at least in part on the fourth confidence values.

4. A method comprising:
accessing first weight data obtained by a first group of one or more weight sensors located at a first side of an inventory location;
determining, based on the first weight data, a first weight change at the first side of the inventory location based on a first weight measurement and a second weight measurement;
accessing second weight data obtained by a second group of one or more weight sensors located at a second side of the inventory location;
determining, based on the second weight data, a second weight change at the second side of the inventory location based on a third weight measurement and a fourth weight measurement;
determining a first set of event hypotheses based at least in part on one or more of the first weight change or the second weight change, wherein each event hypothesis in the first set of event hypotheses describes a possible pick from the inventory location of a possible quantity of a type of item or a possible place to the inventory location of a possible quantity of the type of item; and
determining, using the first weight change and the second weight change as inputs to a first model, a first confidence value for at least one of the event hypotheses in the first set of event hypotheses, wherein:
the first model comprises a joint probability distribution function based on weight values associated with the type of item as measured at both the first side and the second side of the inventory location and based on a location of the type of item as stowed at the inventory location,
the joint probability distribution function accepts as input the first weight change and the second weight change, and
the joint probability distribution function produces as an output the first confidence value indicative of a likelihood that the first weight change and the second weight change are associated with the type of item.

5. The method of claim 4 , further comprising:
selecting one or more event hypotheses in the first set of event hypotheses based at least in part on the first confidence value of the at least one of the event hypotheses in the first set of event hypotheses; and
generating, based on the one or more event hypotheses that are selected, data indicative of one or more of a type of an interaction with the inventory location, a type of item involved in the interaction, or a quantity of the type of item involved in the interaction.

6. The method of claim 4 , wherein the first model further comprises, for the at least one of the event hypotheses in the first set of event hypotheses:
a first probability distribution function based on weight values associated with the type of item as measured at the first side of the inventory location and based on the location of the type of item as stowed at the inventory location, wherein the first probability distribution function accepts as an input the first weight change and produces as an output a second confidence value indicative of a likelihood that the first weight change is associated with the type of item; and
a second probability distribution function based on weight values associated with the type of item as measured at the second side of the inventory location and based on the location of the type of item as stowed at the inventory location, wherein the second probability distribution function accepts as an input the second weight change and produces as an output a third confidence value indicative of a likelihood that the second weight change is associated with the type of item.

7. The method of claim 4 , further comprising:
accessing third weight data that is associated with the type of item and comprises a plurality of weight measurements obtained by the one or more weight sensors;
determining a plurality of detected weight changes based on the plurality of weight measurements;
determining a first probability distribution function based on the plurality of detected weight changes of the type of item,

Assignees

• Amazon Tech Inc

Inventors

• Kumar Dilip
• Bo Liefeng
• Chacko Nikhil
• Crandall Robert
• Desai Nishitkumar Ashokkumar
• Eledath Jayakrishnan Kumar
• Gopal Gopi Prashanth
• Medioni Gerard Guy
• Munger Paul Eugene
• Srivastava Kushagra

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N20/00

  Machine learning · CPC title

• G01G19/4144

  for controlling weight of goods in commercial establishments, e.g. supermarket, P.O.S. systems · CPC title

• G06Q10/087 Primary

  Inventory or stock management, e.g. order filling, procurement or balancing against orders · CPC title