Abstract: This disclosure is generally directed to systems and methods for generating a last mile travel route for a delivery robot. In an example embodiment, an individual captures an image of a frontage of a property. The image may include a package drop-off spot where a package can be dropped off by the delivery robot. The individual may also insert labels and/or illustrations into the image for conveying information such as object descriptors (front door, drop-off spot, etc.), traversable areas (driveway, walkway, etc.), and non-traversable areas (lawn, flower bed, etc.). The image is converted by a computer into a travel map that the delivery robot can use to identify a travel route through the frontage area of the property. The travel map may be provided in various forms such as a semantic map of the frontage area or a bird's eye view rendering of the frontage area.

Abstract: Systems and methods for fleet inspection and maintenance using a robot are provided. The robot may detect a maintenance issue of a vehicle of a fleet of vehicles via one or more sensors, generate a navigation route to a position proximal to the maintenance issue of the vehicle, traverse along the navigation route to the position, and execute a maintenance to rectify the maintenance issue of the vehicle. The robot may include a mobile base removably coupled to a modular platform for performing a maintenance task.

Abstract: This disclosure is generally directed to generating travel route information that is interpretable by a delivery robot for traversing a last mile of a delivery. In an example embodiment, an individual captures a video clip while moving along a travel route that is preferred by the individual for use by the delivery robot. The video clip is converted into a digital route map that the delivery robot uses to reach a package drop-off location on the property. In another example embodiment, an individual captures an image of a portion of the property and appends oral instructions to reach the package drop-off location. The image and the oral instructions are converted into a digital route map for use by the delivery robot. In yet another example embodiment, markers affixed to a surface of a traversable area on the property are used by the delivery robot to reach the package drop-off location.

Abstract: Systems and methods for administering a hazard condition warning during a package delivery operation are provided. The system includes one or more sensors, e.g., radar sensors or cameras, configured to detect a hazard condition in a vicinity of a delivery vehicle and to generate data indicative of the hazard condition. The system further includes an operator sensor configured to detect a location of an operator relative to the delivery vehicle and one or more indicators, e.g., lights, a mobile application installed on a mobile phone, or an HMI of the vehicle, configured to generate an alert corresponding to the hazard condition. A processor of the system may determine whether the hazard condition is present based on the data indicative of the hazard condition, and cause, based on a presence of the hazard condition and the detected location of the operator, the one or more indicators to generate the alert.

Abstract: The present invention extends to methods, systems, and computer program products for classifying time series image data. Aspects of the invention include encoding motion information from video frames in an eccentricity map. An eccentricity map is essentially a static image that aggregates apparent motion of objects, surfaces, and edges, from a plurality of video frames. In general, eccentricity reflects how different a data point is from the past readings of the same set of variables. Neural networks can be trained to detect and classify actions in videos from eccentricity maps. Eccentricity maps can be provided to a neural network as input. Output from the neural network can indicate if detected motion in a video is or is not classified as an action, such as, for example, a hand gesture.

Abstract: This disclosure is generally directed to systems and methods for generating a last mile travel route for a delivery robot. In an example embodiment, an individual captures an image of a frontage of a property. The image may include a package drop-off spot where a package can be dropped off by the delivery robot. The individual may also insert labels and/or illustrations into the image for conveying information such as object descriptors (front door, drop-of spot etc.), traversable areas (driveway, walkway, etc.), and non-traversable areas (lawn, flower bed, etc.). The image is converted by a computer into a travel map that the delivery robot can use to identify a travel, route through the frontage area of the property. The travel map may be provided in various forms such as a semantic map of the frontage area or a bird's eye view rendering of the frontage area.

Abstract: Systems and methods for delivering a package to a curbside receptacle are provided. A delivery window of a vehicle may be aligned with a curbside receptacle. A carousel within the vehicle may be moved, e.g., rotated, within the vehicle to align a target compartment with the delivery window. A conveyor arm slidably coupled to the delivery window is then extended and aligned with the target compartment, and a target package within the target compartment is ejected from the target compartment onto the conveyor arm, e.g., via a plunger mechanism. The conveyor arm then transfers the target package to curbside receptacle.

Abstract: This disclosure is generally directed to generating travel route information that is interpretable by a delivery robot for traversing a last mile of a delivery. In an example embodiment, an individual captures a video clip while moving along a travel route that is preferred by the individual for use by the delivery robot. The video clip is converted into a digital route map that the delivery robot uses to reach a package drop-off location on the property. In another example embodiment, an individual captures an image of a portion of the property and appends oral instructions to reach the package drop-off location. The image and the oral instructions are converted into a digital route map for use by the delivery robot. In yet another example embodiment, markers affixed to a surface of a traversable area on the property are used by the delivery robot to reach the package drop-off location.

Abstract: Disclosed is an end-to-end delivery system that uses standard containers designed for autonomous vehicle (AV) goods delivery, a purpose-built AV cargo management system, a purpose-built robot to load and unload packages, and software to coordinate the various tasks involved. The standard containers may include a hardware locking interface for locking to a carrying robot. The cargo management system may include a programmable conveyor system installed on each floor of a multi-floor cargo space within the AV. For example, the floor may include a roller surface to allow omnidirectional routing of packages. An elevator shaft may be used for receiving and off-loading containers. The software may identify a target container anywhere within the multi-floor cargo space, and determine how to rearrange the containers within a grid in the AV so that the target container may be moved to the elevator shaft for unloading.

Abstract: Systems and methods for fleet inspection and maintenance using a robot are provided. The robot may detect a maintenance issue of a vehicle of a fleet of vehicles via one or more sensors, generate a navigation route to a position proximal to the maintenance issue of the vehicle, traverse along the navigation route to the position, and execute a maintenance to rectify the maintenance issue of the vehicle. The robot may include a mobile base removably coupled to a modular platform for performing a maintenance task.

Abstract: Systems and methods for delivering a package to a curbside receptacle are provided. A delivery window of a vehicle may be aligned with a curbside receptacle. A carousel within the vehicle may be moved. e.g., rotated, within the vehicle to align a target compartment with the delivery window. A conveyor arm slidably coupled to the delivery window is then extended and aligned with the target compartment, and a target package within the target compartment is ejected from the target compartment onto the conveyor arm, e.g., via a plunger mechanism. The conveyor arm then transfers the target package to curbside receptacle.

Abstract: Systems and methods for administering a hazard condition warning during a package delivery operation are provided. The system may include one or more sensors, e.g., radar sensors, ultrasonic sensors, or cameras, configured to detect a hazard condition in a vicinity of a delivery vehicle and to generate data indicative of the hazard condition. The system further includes an operator sensor configured to detect a location of an operator of the delivery vehicle relative to the delivery vehicle, and one or more indicators, e.g., lights, a mobile application installed on a mobile phone, or an HMI of the vehicle, configured to generate an alert corresponding to the hazard condition.

Abstract: The present invention extends to methods, systems, and computer program products for classifying time series image data. Aspects of the invention include encoding motion information from video frames in an eccentricity map. An eccentricity map is essentially a static image that aggregates apparent motion of objects, surfaces, and edges, from a plurality of video frames. In general, eccentricity reflects how different a data point is from the past readings of the same set of variables. Neural networks can be trained to detect and classify actions in videos from eccentricity maps. Eccentricity maps can be provided to a neural network as input. Output from the neural network can indicate if detected motion in a video is or is not classified as an action, such as, for example, a hand gesture.

Abstract: The present invention extends to methods, systems, and computer program products for classifying time series image data. Aspects of the invention include encoding motion information from video frames in an eccentricity map. An eccentricity map is essentially a static image that aggregates apparent motion of objects, surfaces, and edges, from a plurality of video frames. In general, eccentricity reflects how different a data point is from the past readings of the same set of variables. Neural networks can be trained to detect and classify actions in videos from eccentricity maps. Eccentricity maps can be provided to a neural network as input. Output from the neural network can indicate if detected motion in a video is or is not classified as an action, such as, for example, a hand gesture.

Abstract: A lane curvature, a vehicle position offset relative to a lane center line, and a vehicle heading are determined based on image data received from a vehicle camera sensor. An adjusted lane curvature and an adjusted vehicle heading are computed based on a vehicle speed, a vehicle yaw rate and the position offset.

Abstract: The present invention extends to methods, systems, and computer program products for classifying time series image data. Aspects of the invention include encoding motion information from video frames in an eccentricity map. An eccentricity map is essentially a static image that aggregates apparent motion of objects, surfaces, and edges, from a plurality of video frames. In general, eccentricity reflects how different a data point is from the past readings of the same set of variables. Neural networks can be trained to detect and classify actions in videos from eccentricity maps. Eccentricity maps can be provided to a neural network as input. Output from the neural network can indicate if detected motion in a video is or is not classified as an action, such as, for example, a hand gesture.

Abstract: A lane curvature, a vehicle position offset relative to a lane center line, and a vehicle heading are determined based on image data received from a vehicle camera sensor. An adjusted lane curvature and an adjusted vehicle heading are computed based on a vehicle speed, a vehicle yaw rate and the position offset.