Amazon Technologies, Inc.

Chaitanya Mitash

Manikantan Nambi

Vikedo Terhuja

Timothy G Dietz

Gregory Coleman

Tyler W Garaas

Ugur Murat Erdem

Shiyang Lu

Priyesh Shah

First image data is received via a first imaging device that represents an item being manipulated by a robotic arm, and second image data is received via a second imaging device that represents item being manipulated by the robotic arm. Using a machine-learning model to analyze the first image data, a first score is determined that represents whether the item includes a defect. Additionally, using the machine-learning model to analyze the second image data, a second score is determined that represents whether the item includes the defect. Whether the item includes the defect is based on at least one of the first score or the second score, and in such instances, the robotic arm to perform an action.

Identifying defect(s) within item(s) being manipulated by a robotic arm

Robotic end effectors may include a suction cup, one or more grasping fingers, and one or more flexible nails to securely and reliably engage, grasp, and cage items. The grasping fingers may move between open and closed positions to grasp items. In addition, the flexible nails may be coupled to ends of the grasping fingers to facilitate caging of items by the fingers and nails. For example, the flexible nails may bend, flex, or slide under, around, or relative to items and thereby enable grasping and caging of portions of items by the fingers and nails. The nails may be formed of various materials, with various layers or thicknesses of materials, and/or with variable flexibility or friction properties. Further, the nails may be transitioned or actuated between multiple configurations, including a sliding configuration and a grasping configuration.

Robotic end effectors having fingers and nails to grasp and cage items

An item identification and pose determination system is described for use in purposeful selection of items from a collection. The system may be configured to determine a pose of one or more items in a collection of items based on bounding box information associated with the item as well as depth data of the present location of the item. The bounding box is aligned to the depth data using a machine learning algorithm and randomized search algorithm and the pose of the item enables determination of grasp locations for removing the item from the collection and intentionally placing at a destination in a particular pose.

Systems and methods for scalable perception and purposeful robotic picking of items from a collection

An apparatus includes a memory and a hardware processor. The hardware processor determines one or more of a weight of an item, a packaging type of the item, a packaging material of the item, a barcode of the item, a rigidity of the item, or a physical response of the item to being lifted and determines a visual appearance of the item and a shape or size of the item. The hardware processor also compares, using a machine learning model, the determined characteristics of the item to a manifest for the container. The manifest identifies a plurality of items in the container. The hardware processor determines, using the machine learning model, an identity of the item based on comparing the determined characteristics of the item to the manifest.

Growing community of inventors

Waltham, MA, United States of America

Chaitanya Mitash