Abstract: Object picking optimization is disclosed, including: receiving an image of a target object; inputting the image in a machine learning model that is configured to output a pick probability heat map corresponding to the target object; selecting a pick location on the target object based at least in part on the pick probability heat map corresponding to the target object; and causing a diverting mechanism to perform a pick operation on the target object based at least in part on the pick location.

Abstract: Pick quality determination is disclosed, including: using a pressure meter to sense a pressure associated with an airflow through a gripper mechanism of a diverting mechanism over time during a pick operation on a target object; storing the sensed pressure associated with the airflow through the gripper mechanism over time as a pressure sequence associated with the pick operation on the target object; and correlating the pressure sequence with representative pressure sequences associated with corresponding pick quality types to determine whether the pick operation on the target object was successful or not.

Abstract: Cloud and facility-based machine learning for sorting facilities is disclosed, including: obtaining a machine learning model associated with a domain associated with materials to be sorted at a first sorting facility; and generating a modified machine learning model by training the machine learning model using data obtained from the first sorting facility.

Abstract: Using machine learning to recognize variant objects is disclosed, including: identifying an object as a variant of an object type by inputting sensed data associated with the object into a modified machine learning model corresponding to the variant of the object type, wherein the modified machine learning model corresponding to the variant of the object type is generated using a machine learning model corresponding to the object type; and generating a control signal to provide to a sorting device that is configured to perform a sorting operation on the object, wherein the sorting operation on the object is determined based at least in part on the variant of the object type associated with the object.

Abstract: Maintaining a data structure corresponding to a target object is disclosed, including: determining that an identified target object from a sensed signal is a new target object, wherein the sensed signal is generated at a sorting facility; generating a new data structure corresponding to the new target object; and updating the new data structure with an attribute associated with the new target object, wherein the attribute is derived from the sensed signal associated with the new target object.

Abstract: Object material type identification using multiple types of sensors is disclosed, including: obtaining a machine learning model, wherein the machine learning model has been trained using training data comprising vision sensor data on a set of objects, and wherein the vision sensor data on the set of objects is associated with material characteristic labels that are determined based at least in part on non-vision sensor data on the set of objects; obtaining a vision sensor signal corresponding to an object; and using the machine learning model and the vision sensor signal to determine a material characteristic type associated with the object.

Abstract: Heterogenous material sorting is disclosed, including: identifying a first target object associated with a first object type on a surface based at least in part on a first sensed signal; providing a first control signal to a first sorting device to cause the first sorting device to remove the first target object from the surface, wherein the first sorting device is configured to manipulate objects associated with the first object type; identifying a second target object associated with a second object type on the surface based at least in part on a second sensed signal; and providing a second control signal to a second sorting device to cause the second sorting device to remove the second target object from the surface, wherein the second sorting device is configured to manipulate objects associated with the second object type.

Abstract: Object path planning in a sorting facility is disclosed, including: obtain data describing a trajectory associated with a target object; generate a control signal for a sorting device to perform a sorting operation on the target object based at least in part on the trajectory associated with the target object; and provide the control signal to the sorting device, wherein the sorting device is configured to execute the control signal with respect to the target object.