Amazon Technologies, Inc.

Zheng Wang

Shao-Wen Yang

Xingang Guo

Qianchao Ban

Jin Li

Steven Scott Larson

Hoshgeldy Tachmuradov

Karthik Ram Srinivasan

Matthew Kenneth Bellamy

HebaAllah Aly AbdElHalim Aly Ismail

Bettina Cornelia Sehorsch

Hebaallah Aly Abdelhalim Aly Ismail

Ruth Ravichandran

Yang Pan

Yee Yan Cheng

Andrey Li

David Hung Huynh

Various embodiments are disclosed for providing machine learning routines with peripheral device data to infer driver activity and location. Peripheral device data may be collected on a peripheral device having a machine learning routine executing thereon to infer driver activity and perform improved estimation of driver location. Using driver activity and location estimation, contextually relevant delivery workflow assistance may be automatically provided to a delivery driver or other individual without requiring manual input, thereby improving driver safety and operational efficiency.

Machine learning item delivery using sensor data

A system can be configured to receive and analyze an image captured by a user device at a delivery location. The system can utilize local algorithms tailored to the delivery location to confirm that the delivery location is the intended delivery location for a package. The system can utilize global algorithms to generate three-dimensional models of the delivery environment and an augmented reality interface providing delivery instructions to the deliverer. The delivery instructions include indications of approved delivery locations and rejected delivery locations associated with the general delivery location of the customer. The delivery instructions and the user device can be utilized to ensure package placement for secure delivery to a customer.

Location-based package drop-off instructions

Disclosed are various embodiments for generating location confirmation models using crowdsourced wireless fingerprints. Wireless fingerprints can be generated and associated with a given location during events that use proximity to specific locations. The wireless fingerprints can be processed to generate location confirmation models that can be used for location confirmation. Periodically, the collected wireless fingerprints can be analyzed and compared to previously collected wireless fingerprints to detect a change in a wireless infrastructure at the given location. Upon determining that a previously generated location confirmation model is invalid according to a level of significance of the change, the outdated wireless fingerprints can be identified and removed from storage or otherwise ignored for future models. An updated location confirmation model can be generated using the up-to-date wireless fingerprints.

Location confirmation using crowdsourced wireless fingerprints

A delivery confirmation of a package at a delivery location may be determined based on a communication protocol between a user device and a package device associated with the package. A user can order an item and anticipate a delivery of the item to a delivery location associated with the user. A delivery device can determine that it has entered a geographic region associated with the delivery location. The delivery device and/or a user device associated with the delivery location can determine, using the communication protocol, that the delivery device is within a range of the user device. Then, the user device can receive a package device identifier associated with the package device, determine that the package device is within the range of the user device, and determine a delivery confirmation of the package.

Beacon-based delivery confirmation

Disclosed are various embodiments for estimating missing wireless attributes of a wireless fingerprint generated for a given location and confirming that the given location is a correct location. A collection of wireless fingerprint samples for the given location can be used to train an attribute prediction model that can estimate missing attributes in wireless fingerprints. For example, a wireless fingerprint may be incomplete as the fingerprint may fail to include attribute data for some AP devices associated with the given location. The attribute prediction model can be used to reconstruct wireless fingerprints to include the missing attributes. The reconstructed wireless fingerprints can then be used to train a location confirmation model which can be used for accurate location confirmation.

Wireless fingerprint reconstruction for accurate location confirmation

Growing community of inventors

Sammamish, WA, United States of America

Zheng Wang