Amazon Technologies, Inc.

Nicholas A Smith

Charles C Linton

Oliver Christoph Purwin

Lars Petter Johnsen

Techniques for mobile robots performing self-diagnostic fiducial classification are described herein. For example, a mobile robot can determine that an expected fiducial marker has not been detected as the mobile robot navigates via fiducial markers on a surface. In response, the mobile robot can generate an input to a machine learning model that can be executed on the mobile robot. The input can include images of the surface captured by the mobile robot. The machine learning model may be trained to generate an output comprising a classification of image data of the images based at least in part on the input. The mobile robot can determine an error event corresponding to the expected fiducial marker based at least in part on the classification of image data received from the machine learning model. The mobile robot can output a notification indicating the error event corresponding to the expected fiducial marker.

Self-diagnostic fiducial classifier for mobile robots

In various examples, systems and methods for qualification of robotic firmware are described. In various examples, first firmware or first hardware may be deployed on a first set of robots and second firmware or second hardware may be deployed on a second set of robots. The first firmware may generate first log data and the second firmware may generate second log data during operation of the first and second set of robots during a first time period. A first value of a first performance metric may be determined using the first log data and a second value of the first performance metric may be determined using the second log data. The second firmware or the second hardware may be modified based at least in part on the first value and the second value.

Growing community of inventors

Medford, MA, United States of America

Nicholas A Smith