Abstract: Implementations described herein relate to training and refining robotic control policies using imitation learning techniques. A robotic control policy can be initially trained based on human demonstrations of various robotic tasks. Further, the robotic control policy can be refined based on human interventions while a robot is performing a robotic task. In some implementations, the robotic control policy may determine whether the robot will fail in performance of the robotic task, and prompt a human to intervene in performance of the robotic task. In additional or alternative implementations, a representation of the sequence of actions can be visually rendered for presentation to the human can proactively intervene in performance of the robotic task.

Abstract: A driving circuit, a driving method and a microfluidic substrate are provided. The driving circuit includes a first switching unit, a second switching unit, a reset unit, a first capacitor, and a second capacitor. In a first stage of a driving process of the driving circuit, the first switching unit is turned on, a first voltage signal is transmitted to a first node, the second switching unit is turned on, a second voltage signal is input to an output terminal of the driving circuit, and the driving circuit outputs an AC signal. In a second stage of the driving process, the first switching unit is turned off, the valid signal output by the second scan signal terminal controls the reset unit to be turned on, a third voltage signal is input to the output terminal of the driving circuit for reset, and the driving circuit outputs a DC signal.

Abstract: Implementations are directed to training a machine learning model that, once trained, is used in performance of robotic grasping and/or other manipulation task(s) by a robot. The model can be trained using simulated training examples that are based on simulated data that is based on simulated robot(s) attempting simulated manipulations of various simulated objects. Portion(s) of the model can also be trained based on real training examples that are based on data from real-world physical robots attempting manipulations of various objects. The simulated training examples can be utilized to train the model to predict an output that can be utilized in a particular task—and the real training examples used to adapt at least a portion of the model to the real-world domain can be tailored to a distinct task. In some implementations, domain-adversarial similarity losses are determined during training, and utilized to regularize at least portion(s) of the model.

Abstract: A gene sequencing structure, a gene sequencing device and a gene sequencing method are provided. The gene sequencing structure includes a substrate, a thin-film transistor array layer located on the substrate and including thin-film transistors that include a first electrode, and a second electrode; an ion-sensitive layer located on a side of the semiconductor layer away from the substrate; a micro-hole layer located on a side of the ion-sensitive layer away from the substrate, including a through-hole passing through the micro-hole layer, at least partially overlapping the semiconductor layer, and used for receiving a to-be-tested single-stranded nucleic acid inside; a conductive structure, located on a side of the layer away from the substrate and electrically connected to the first electrode or the second electrode; and a detection chip, located on a side of the conductive structure away from the substrate and electrically connected to the conductive structure.

Abstract: Training and/or using a machine learning model for locomotion control of a robot, where the model is decoupled. In many implementations, the model is decoupled into an open loop component and a feedback component, where a user can provide a desired reference trajectory (e.g., a symmetric sine curve) as input for the open loop component. In additional and/or alternative implementations, the model is decoupled into a pattern generator component and a feedback component, where a user can provide controlled parameter(s) as input for the pattern generator component to generate pattern generator phase data (e.g., an asymmetric sine curve). The neural network model can be used to generate robot control parameters.

Abstract: A detection device and a detection method are provided. The detection device includes at least one detection unit. The detection unit includes a first transistor, a second transistor, a third transistor and a fourth transistor that are electrically connected to each other, a gate is disposed above a channel of each of the first transistor, the second transistor, and the third transistor, and an ion-sensitive membrane is covered above a channel of the fourth transistor. The detection device also includes a first voltage signal terminal, a second voltage signal terminal, and a third voltage signal terminal. Further, the detection device includes a first power supply terminal, a first potential output terminal, a second potential output terminal, and a second power supply terminal.

Abstract: Generating edge-depth values for an object, utilizing the edge-depth values in generating a 3D point cloud for the object, and utilizing the generated 3D point cloud for generating a 3D bounding shape (e.g., 3D bounding box) for the object. Edge-depth values for an object are depth values that are determined from frame(s) of vision data (e.g., left/right images) that captures the object, and that are determined to correspond to an edge of the object (an edge from the perspective of frame(s) of vision data). Techniques that utilize edge-depth values for an object (exclusively, or in combination with other depth values for the object) in generating 3D bounding shapes can enable accurate 3D bounding shapes to be generated for partially or fully transparent objects. Such increased accuracy 3D bounding shapes directly improve performance of a robot that utilizes the 3D bounding shapes in performing various tasks.

Abstract: Implementations are directed to training a machine learning model that, once trained, is used in performance of robotic grasping and/or other manipulation task(s) by a robot. The model can be trained using simulated training examples that are based on simulated data that is based on simulated robot(s) attempting simulated manipulations of various simulated objects. Portion(s) of the model can also be trained based on real training examples that are based on data from real-world physical robots attempting manipulations of various objects. The simulated training examples can be utilized to train the model to predict an output that can be utilized in a particular task—and the real training examples used to adapt at least a portion of the model to the real-world domain can be tailored to a distinct task. In some implementations, domain-adversarial similarity losses are determined during training, and utilized to regularize at least portion(s) of the model.

Abstract: A microfluidic chip and a fabrication method of the microfluidic chip are provided. The microfluidic chip includes an array substrate, and a hydrophobic layer disposed on a side of the array substrate. The hydrophobic layer includes at least one through-hole, and a through-hole of the at least one through-hole penetrates through the hydrophobic layer along a direction perpendicular to a plane of the array substrate. The microfluidic chip also includes at least one hydrophilic structure. A hydrophilic structure of the at least one hydrophilic structure is disposed in the through-hole.

Abstract: Implementations relate to training a point cloud prediction model that can be utilized to process a single-view two-and-a-half-dimensional (2.5D) observation of an object, to generate a domain-invariant three-dimensional (3D) representation of the object. Implementations additionally or alternatively relate to utilizing the domain-invariant 3D representation to train a robotic manipulation policy model using, as at least part of the input to the robotic manipulation policy model during training, the domain-invariant 3D representations of simulated objects to be manipulated. Implementations additionally or alternatively relate to utilizing the trained robotic manipulation policy model in control of a robot based on output generated by processing generated domain-invariant 3D representations utilizing the robotic manipulation policy model.

Abstract: Techniques are disclosed that enable training a meta-learning model, for use in causing a robot to perform a task, using imitation learning as well as reinforcement learning. Some implementations relate to training the meta-learning model using imitation learning based on one or more human guided demonstrations of the task. Additional or alternative implementations relate to training the meta-learning model using reinforcement learning based on trials of the robot attempting to perform the task. Further implementations relate to using the trained meta-learning model to few shot (or one shot) learn a new task based on a human guided demonstration of the new task.

Abstract: Mitigating the reality gap through training and utilization of at least one difference model. The difference model can be utilized to generate, for each of a plurality of instances of simulated state data generated by a robotic simulator, a corresponding instance of modified simulated state data. The difference model is trained so that a generated modified instance of simulated state data is closer to “real world data” than is a corresponding initial instance of simulated state data. Accordingly, the difference model can be utilized to mitigate the reality gap through modification of initially generated simulated state data, to make it more accurately reflect what would occur in a real environment. Moreover, the difference representation from the difference model can be used as input to the control policy to adapt the control learned from simulator to the real environment.

Abstract: Active utilization of a robotic simulator in control of one or more real world robots. A simulated environment of the robotic simulator can be configured to reflect a real world environment in which a real robot is currently disposed, or will be disposed. The robotic simulator can then be used to determine a sequence of robotic actions for use by the real world robot(s) in performing at least part of a robotic task. The sequence of robotic actions can be applied, to a simulated robot of the robotic simulator, to generate a sequence of anticipated simulated state data instances. The real robot can be controlled to implement the sequence of robotic actions. The implementation of one or more of the robotic actions can be contingent on a real state data instance having at least a threshold degree of similarity to a corresponding one of the anticipated simulated state data instances.

Abstract: Microfluidic device and application method thereof are provided. The microfluidic device includes a first substrate and a second substrate that are oppositely arranged along a first direction; and a first storage box and a second storage box that are oppositely arranged along the first direction. The first direction is a thickness direction of the microfluidic device. The first storage box includes a first storage cavity and a first opening communicating with the first storage cavity, and the first substrate is fixed in the first storage cavity. The second storage box includes a second storage cavity and a second opening communicating with the second storage cavity, and the second substrate is fixed in the second storage cavity. Along the first direction, the first opening is arranged opposite to the second opening and the first storage box is nested with the second storage box.

Abstract: Provided is a microfluidic chip. The microfluidic chip includes a first substrate and a second substrate disposed opposite to each other and drive electrodes, first sensing electrodes and second sensing electrodes disposed on a side of the first substrate. A microfluidic channel is formed between the first substrate and the second substrate and configured to accommodate at least one droplet. Different drive voltage signals are applied to adjacent drive electrodes, so as to drive the at least one droplet to move. Detection signals are applied to the first sensing electrodes and the second sensing electrodes, and a position of the at least one droplet is determined according to a change in capacitance between one first sensing electrode and an electrode corresponding thereto and a change in capacitance between one second sensing electrode and an electrode corresponding thereto when the at least one droplet flows by.

Abstract: Microfluidic substrate, microfluidic device, and driving method thereof are provided. The microfluidic substrate includes a plurality of detection units arranged in an array. A detection unit of the plurality of detection units at least includes a first switch transistor, a second switch transistor, a drive electrode, and a photosensitive element. The microfluid substrate includes a base; a transistor array layer on a side of the base, first switch transistors and second switch transistors being on the transistor array layer; a photosensitive element array layer on a side of the transistor array layer away from the substrate, photosensitive elements being on the photosensitive element array layer; a first electrode layer on a side of the photosensitive element array layer away from the base; and a second electrode layer on a side of the first electrode layer away from the base.

Abstract: Active utilization of a robotic simulator in control of one or more real world robots. A simulated environment of the robotic simulator can be configured to reflect a real world environment in which a real robot is currently disposed, or will be disposed. The robotic simulator can then be used to determine a sequence of robotic actions for use by the real world robot(s) in performing at least part of a robotic task. The sequence of robotic actions can be applied, to a simulated robot of the robotic simulator, to generate a sequence of anticipated simulated state data instances. The real robot can be controlled to implement the sequence of robotic actions. The implementation of one or more of the robotic actions can be contingent on a real state data instance having at least a threshold degree of similarity to a corresponding one of the anticipated simulated state data instances.

Abstract: Provided is a microfluidic device. The microfluidic device includes a first substrate and a second substrate disposed opposite to each other. A cavity is formed between the first substrate and the second substrate and configured to accommodate liquid. The first substrate includes multiple drive electrodes and multiple first electrodes, and the drive electrodes are disposed on a side of the first electrodes facing the second substrate. At least one of the drive electrodes includes at least one opening, and the at least one opening, along a direction perpendicular to a plane where the first substrate is located, penetrates the drive electrode where the at least one opening is located. An orthographic projection of at least one first electrode on the plane where the first substrate is located covers at least an orthographic projection of one opening on the plane where the first substrate is located.

Abstract: Provided a microfluidic pixel driving circuit includes n boost modules, where each boost module includes a capacitor and a write unit, and n is a positive integer greater than or equal to 2; a first terminal of a first capacitor is electrically connected to a fixed potential line, a second terminal of the first capacitor is electrically connected to a pixel electrode, and the first capacitor is used for storing a voltage of the pixel electrode; a first write unit is configured to write a first data signal to the pixel electrode according to an enable level of a first scan signal; a first terminal of a second capacitor is electrically connected to the pixel electrode; and a second write unit is configured to write a second data signal to a second terminal of the second capacitor according to an enable level of a second scan signal.

Abstract: Training a machine learning model (e.g., a neural network model such as a convolutional neural network (CNN) model) so that, when trained, the model can be utilized in processing vision data (e.g., from a vision component of a robot), that captures an object, to generate a rich object-centric embedding for the vision data. The generated embedding can enable differentiation of even subtle variations of attributes of the object captured by the vision data.