A soft gripper including tentacles, each tentacle includes lower and upper members connected by a connector. Each member includes guide discs, and each guide disc includes a ring with passthrough holes, and a spacer located in a donut hole of the ring with passthrough holes, the passthrough holes collectively define cable pathways. The connector includes a center thru-hole and transfer channels. Cables have proximal ends attached to actuators and extend through apertures of a baseplate located at a proximal end of the lower member. A set of lower cables extend through the lower ring passthrough holes to couple to a distal lower guide disc. A set of upper cables extend through the lower spacer passthrough holes, through the transfer channels to the upper ring passthrough holes to couple to a distal upper guide ring, and an end cap is attached to the distal end of the upper member.

A robot includes: an arm; a hand including a first finger and a second finger, wherein the first finger includes a first sensor and the second finger includes a second sensor; and a processor configured to: based on sensing an object through the first sensor while the robot is moving to grip the object, activate the second sensor, receive, from the first sensor and the second sensor, distance information including a plurality of pairs of distance values, wherein each respective pair of distance values of the plurality of pairs of distance values includes a first distance between the first sensor and the object and a second distance between the second sensor and the object, and each respective pair of distance values corresponds to a respective position of the hand relative to the object, and control the first finger and the second finger to grip the object based on the distance information.

A technique for training a neural network, including generating a plurality of input vectors based on a first plurality of task demonstrations associated with a first robot performing a first task in a simulated environment, wherein each input vector included in the plurality of input vectors specifies a sequence of poses of an end-effector of the first robot, and training the neural network to generate a plurality of output vectors based on the plurality of input vectors. Another technique for generating a task demonstration, including generating a simulated environment that includes a robot and at least one object, causing the robot to at least partially perform a task associated with the at least one object within the simulated environment based on a first output vector generated by a trained neural network, and recording demonstration data of the robot at least partially performing the task within the simulated environment.

A first method comprising: predicting a scene of an environment using a model of the environment and based on a first scene of the environment obtained from sensors observing scenes of the environment; comparing the predicted scene with an observed scene from the sensors; and performing an action based on differences determined between the predicted scene and the observed scene. A second method comprising applying a vibration stimuli on an object via a computer-controlled component; obtaining a plurality of images depicting the object from a same viewpoint, captured during the application of the vibration stimuli. The second method further comprising comparing the plurality of images to detect changes occurring in response to the application of the vibration stimuli, which changes are attributed to a change of a location of a boundary of the object; and determining the boundary of the object based on the comparison.

Systems and methods are described, and an example system includes a transport bin configured to carry a baggage item and having spatial reference frame marking detectable by electromagnetic scan and by machine vision. The system includes a robotic arm apparatus at an inspection area, and includes a switched path baggage conveyor that, responsive to electromagnetic scan detection of an object-of-interest (OOI) within the baggage item, conveys the transport bin to the inspection area. The electromagnetic scan generates OOI geometric position information indicating geometric position of the OOI relative to the spatial reference frame marking. The robotic arm apparatus, responsive to receiving the transport bin, uses machine vision to detect orientation of the spatial reference frame marking, then translates OOI geometric position information to local reference frame, for robotic opening of the baggage item, and robotic accessing and contact swab testing on the OOI.

Systems and techniques for target object retrieval may include or utilize an image capture device, and a task planner. The image capture device may receive an image of an environment including identified objects. The task planner may determine potential actions, calculate a probability of success of achieving a desired goal for each of the potential actions based on an action prediction model, the corresponding potential action, a current state of the environment, any previously taken action, and the desired goal, select a potential action associated with the highest calculated probability of success, and simulate a subsequent state based on the selected potential action and a dynamic prediction model. The potential actions may be associated with an identified object of the identified objects and an operation to be performed on the identified object.

A simulation apparatus includes a processor that executes a simulation of a control program executed on a controller. The controller controls motion of a machine that handles an object. The processor includes: a motion control device that controls motion of a virtual machine based on a motion command to move the virtual machine in a virtual space, with the virtual machine corresponding to the machine; a determination device that determines whether a volume of a region, where a work space in which the virtual machine works overlaps with the virtual object, is equal to or greater than a predetermined reference value, the virtual object being handled by the virtual machine and corresponding to the object; and a follow-up device that makes the virtual object follow the motion of the virtual machine based on the motion command when the volume is equal to or greater than the reference value.

Provided is a fruit quality inspecting and sorting appliance. The fruit quality inspecting and sorting appliance includes: a conveying module; a weighing module, which cooperates with the conveying module to convey weighed fruits through the conveying module; an internal quality inspection module, which cooperates with the conveying module and performs an internal quality inspection to the weighed fruits; an external quality inspection module, which cooperates with the conveying module and performs an external quality inspection to the fruits after the internal quality inspection; a sorting module, which cooperates with the conveying module, and sorts the fruits passing through the weighing module, the internal quality inspection module and the external quality inspection module; and a control module electrically connected with the conveying module, the weighing module, the internal quality inspection module, the external quality inspection module and the sorting module.

Methods and apparatus for determining a grasp strategy to grasp an object with a gripper of a robotic device are described. The method comprises generating a set of grasp candidates to grasp a target object, wherein each of the grasp candidates includes information about a gripper placement relative to the target object, determining, for each of the grasp candidates in the set, a grasp quality, wherein the grasp quality is determined using a physical-interaction model including one or more forces between the target object and the gripper located at the gripper placement for the respective grasp candidate, selecting, based at least in part on the determined grasp qualities, one of the grasp candidates, and controlling the robotic device to attempt to grasp the target object using the selected grasp candidate.

Some robotic arms may include vacuum-based grippers. Detecting the seal quality between each vacuum assembly of the gripper and a grasped object may enable reactivation of some vacuum assemblies, thereby improving the grasp. One embodiment of a method may include activating each of a plurality of vacuum assemblies of a robotic gripper by supplying a vacuum to each vacuum assembly, determining, for each of the activated vacuum assemblies, a first respective seal quality of the vacuum assembly with a first grasped object, deactivating one or more of the activated vacuum assemblies based, at least in part, on the first respective seal qualities, and reactivating each of the deactivated vacuum assemblies within a reactivation interval.