Systems and methods monitor a workspace for safety purposes using sensors distributed about the workspace. The sensors are registered with respect to each other, and this registration is monitored over time. Occluded space as well as occupied space is identified, and this mapping is frequently updated.

A detection system includes a first detection apparatus that detects an object that is being moved, within a predetermined detection region, a number of times, a work-data creation device that creates, every time the first detection apparatus detects the object, work data having a first data element that indicates at least a position of the object obtained by the first detection apparatus and a second data element that includes at least an index related to the object and obtained at the time of the detection, and a work-data storage unit that stores the work data created by the work-data creation means. The work-data storage unit selects, as the work data that should be stored, one of the work data that is newly created for the object and the work data for the object that has been stored by the work-data storage unit, on the basis of the index.

Provided is a control apparatus including: an assessment unit configured to assess whether a relevant element is represented by environment information acquired from a sensor or not; and an environment information setting unit configured to, in a case where the assessment unit has assessed that the relevant element is represented by the environment information, switch the environment information to acquired-in-advance environment information in which the relevant element is not included.

Methods, systems, and apparatus, including computer programs encoded on computer storage media for evaluating robot learning. In some implementations, a system receives classification examples from a plurality of remote devices over a communication network. The classification examples can include (i) a data representation generated by a remote device based on sensor data captured by the remote device and (ii) a classification corresponding to the data representation. The system assigns quality scores to the classification examples based on a level of similarity of the data representations with other data representations. The system selects a subset of the classification examples based on the quality scores assigned to the classification examples. The system trains a machine learning model using the selected subset of the classification examples.

A robotic manipulation planning system, including at least one processor; and a non-transitory computer-readable storage medium including instructions that, when executed by the at least one processor, cause the at least one processor to: process perception data to detect known, familiar, and unknown objects to generate manipulation candidates; filter manipulation candidates against constraints to reduce the manipulation candidates; and determine quality metrics for the reduced manipulation candidates using a soft-body simulation technique.

Embodiments of the present invention determine the configuration of a workpiece and whether it is actually being handled by a monitored piece of machinery, such as a robot. The problem solved by the invention is especially challenging in real-world factory environments because many objects, most of which are not workpieces, may be in proximity to the machinery. Accordingly, embodiments of the invention utilize semantic understanding to distinguish between workpieces that may become associated with the robot and other objects (and humans) in the workspace that will not, and detect when the robot is carrying a workpiece.

Systems and methods monitor a workspace for safety purposes using sensors distributed about the workspace. The sensors are registered with respect to each other, and this registration is monitored over time. Occluded space as well as occupied space is identified, and this mapping is frequently updated. Based on the mapping, a constrained motion plan of machinery can be generated to ensure safety.

A point cloud information acquiring unit for acquiring point cloud information of a region including at least one of the picking hand and the work at a time which is at least one of (i) when the picking hand picks the work at the first position, (ii) when the picking hand moves the work out of the first container, (iii) when the picking hand moves the work in the second container and (iv) when the picking hand arranges the work at the second position; and an abnormality detection unit for detecting an abnormality based on the point cloud information acquired by the point cloud information acquiring unit are included.

Methods, systems, and apparatus, including computer programs encoded on computer storage media for evaluating robot learning. In some implementations, one or more computers receive object classification examples from a plurality of robots. Each object classification example includes (i) an embedding that a robot generated using a machine learning model, and (ii) an object classification corresponding to the embedding. The object classification examples are evaluated based on a similarity of the received embeddings with respect to other embeddings. A subset of the object classification examples is selected based on the evaluation of the quality of the embeddings. The subset of the object classification examples is distributed to the robots in the plurality of robots.

A robotic pallet includes a plurality of composable posts, each of the plurality of composable posts having a processor and a memory. The plurality of composable posts are selectively arranged and interconnected to form the robotic pallet, with one or more locks positioned on the robotic pallet, to releasably securable to a product pallet. In some versions, the robotic pallet is supported by a platform which may be formed by composable posts, and may include a number of robotic manipulation arms.