Abstract: Features are disclosed for an end effector for automated identification and handling of an object. The end effector includes an end effector that can be positioned over a pick point of an overpackage in which a desired object is location using sensors. Using the location information, the end effector can identify a path to the pick point and detect whether the pick point is engaged by detecting environmental changes at the end effector.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network. One of the methods includes maintaining a replay memory that stores trajectories generated as a result of interaction of an agent with an environment; and training an action selection neural network having policy parameters on the trajectories in the replay memory, wherein training the action selection neural network comprises: sampling a trajectory from the replay memory; and adjusting current values of the policy parameters by training the action selection neural network on the trajectory using an off-policy actor critic reinforcement learning technique.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection policy neural network, wherein the action selection policy neural network is configured to process an observation characterizing a state of an environment to generate an action selection policy output, wherein the action selection policy output is used to select an action to be performed by an agent interacting with an environment. In one aspect, a method comprises: obtaining an observation characterizing a state of the environment subsequent to the agent performing a selected action; generating a latent representation of the observation; processing the latent representation of the observation using a discriminator neural network to generate an imitation score; determining a reward from the imitation score; and adjusting the current values of the action selection policy neural network parameters based on the reward using a reinforcement learning training technique.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network used to select actions to be performed by an agent interacting with an environment. One of the methods includes: obtaining data identifying a set of trajectories, each trajectory comprising a set of observations characterizing a set of states of the environment and corresponding actions performed by another agent in response to the states; obtaining data identifying an encoder that maps the observations onto embeddings for use in determining a set of imitation trajectories; determining, for each trajectory, a corresponding embedding by applying the encoder to the trajectory; determining a set of imitation trajectories by applying a policy defined by the neural network to the embedding for each trajectory; and adjusting parameters of the neural network based on the set of trajectories, the set of imitation trajectories and the embeddings.

Abstract: Systems, methods, and apparatus, including computer programs encoded on a computer storage medium, for selecting an actions from a set of actions to be performed by an agent interacting with an environment. In one aspect, the system includes a dueling deep neural network. The dueling deep neural network includes a value subnetwork, an advantage subnetwork, and a combining layer. The value subnetwork processes a representation of an observation to generate a value estimate. The advantage subnetwork processes the representation of the observation to generate an advantage estimate for each action in the set of actions. The combining layer combines the value estimate and the respective advantage estimate for each action to generate a respective Q value for the action. The system selects an action to be performed by the agent in response to the observation using the respective Q values for the actions in the set of actions.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network. One of the methods includes maintaining a replay memory that stores trajectories generated as a result of interaction of an agent with an environment; and training an action selection neural network having policy parameters on the trajectories in the replay memory, wherein training the action selection neural network comprises: sampling a trajectory from the replay memory; and adjusting current values of the policy parameters by training the action selection neural network on the trajectory using an off-policy actor critic reinforcement learning technique.

Abstract: Systems, methods, and apparatus, including computer programs encoded on a computer storage medium, for selecting an actions from a set of actions to be performed by an agent interacting with an environment. In one aspect, the system includes a dueling deep neural network. The dueling deep neural network includes a value subnetwork, an advantage subnetwork, and a combining layer. The value subnetwork processes a representation of an observation to generate a value estimate. The advantage subnetwork processes the representation of the observation to generate an advantage estimate for each action in the set of actions. The combining layer combines the value estimate and the respective advantage estimate for each action to generate a respective Q value for the action. The system selects an action to be performed by the agent in response to the observation using the respective Q values for the actions in the set of actions.

Abstract: A neural network control system for controlling an agent to perform a task in a real-world environment, operates based on both image data and proprioceptive data describing the configuration of the agent. The training of the control system includes both imitation learning, using datasets generated from previous performances of the task, and reinforcement learning, based on rewards calculated from control data output by the control system.

Abstract: Systems, methods, and apparatus, including computer programs encoded on a computer storage medium, for selecting an actions from a set of actions to be performed by an agent interacting with an environment. In one aspect, the system includes a dueling deep neural network. The dueling deep neural network includes a value subnetwork, an advantage subnetwork, and a combining layer. The value subnetwork processes a representation of an observation to generate a value estimate. The advantage subnetwork processes the representation of the observation to generate an advantage estimate for each action in the set of actions. The combining layer combines the value estimate and the respective advantage estimate for each action to generate a respective Q value for the action. The system selects an action to be performed by the agent in response to the observation using the respective Q values for the actions in the set of actions.

Abstract: The present invention discloses a mobile terminal-based photo-taking method and said mobile terminal, the method comprising: after determining that the camera is working in the window view mode before an application needs to call the camera to shoot, the mobile terminal sets the view window of the camera in a small screen display mode on the application interface of the application, such that, when a click action on the view window is detected, the camera is called to shoot a picture. The operations are simple and convenient, and the user experience is improved.

Abstract: Systems, methods, and apparatus, including computer programs encoded on a computer storage medium, for selecting an actions from a set of actions to be performed by an agent interacting with an environment. In one aspect, the system includes a dueling deep neural network. The dueling deep neural network includes a value subnetwork, an advantage subnetwork, and a combining layer. The value subnetwork processes a representation of an observation to generate a value estimate. The advantage subnetwork processes the representation of the observation to generate an advantage estimate for each action in the set of actions. The combining layer combines the value estimate and the respective advantage estimate for each action to generate a respective Q value for the action. The system selects an action to be performed by the agent in response to the observation using the respective Q values for the actions in the set of actions.

Abstract: A continuous process for producing hemicellulose sugars from a biomass extraction liquor is provided. A system is configured for continuously producing hemicellulose sugars and/or hemicellulose derivatives from a biomass extraction liquor, the system comprising at least a first hydrolysis reactor and a second hydrolysis reactor. Each of the hydrolysis reactors is in switchable communication with (i) an operating feed stream of a biomass extraction liquor containing water, hemicellulose oligomers, and dissolved or suspended lignin, and (ii) a cleaning feed stream of a cleaning agent selected from the group consisting of steam, an alkaline solution, an organic solvent, and combinations thereof. The cleaning agent dissolves precipitated lignin formed from the lignin under the hydrolysis reaction conditions.

Abstract: A continuous process for producing hemicellulose sugars from a biomass extraction liquor is provided. A system is configured for continuously producing hemicellulose sugars and/or hemicellulose derivatives from a biomass extraction liquor, the system comprising at least a first hydrolysis reactor and a second hydrolysis reactor. Each of the hydrolysis reactors is in switchable communication with (i) an operating feed stream of a biomass extraction liquor containing water, hemicellulose oligomers, and dissolved or suspended lignin, and (ii) a cleaning feed stream of a cleaning agent selected from the group consisting of steam, an alkaline solution, an organic solvent, and combinations thereof. The cleaning agent dissolves precipitated lignin formed from the lignin under the hydrolysis reaction conditions.

Abstract: The invention provides a method for purifying a biomass hydrolysate comprising sugars and suspended particles, comprising centrifuging the biomass hydrolysate, thermally treating the centrifuged hydrolysate to chemically or physically agglomerate the suspended particles, and filtering the thermally treated hydrolysate to remove agglomerated suspended particles, thereby generating a purified hydrolysate (sugar syrup). The sequence of steps may be varied. Biomass hydrolysates may be provided from a wide variety of processes. Surprisingly, a 20-fold improvement in sugar purity (total suspended solids content) is demonstrated experimentally, compared to prior methods.

Abstract: The present invention provides a process for fractionating lignocellulosic biomass, comprising: digesting a biomass feedstock in the presence of a solvent for lignin, an acid, and water, to produce cellulose-rich solids; separating and washing the cellulose-rich solids with a wash solvent; washing the cellulose-rich solids with water, to generate washed cellulose-rich solids and a wash liquor comprising fines, wherein the wash liquor is introduced to or in contact with a classifier to remove the fines; and separating the fines and recycling the remaining water. The classifier may include a screen with mesh size in the range of 10 to 500, such as 200. The washed cellulose-rich solids will typically have a lower Kappa number (lignin content) and ash content compared to cellulose-rich solids from a process without a classifier that removes fines.

Abstract: The present invention provides a signal processing method and device for the fiber-optic gyroscope, which can effectively expand the dynamic range of the fiber-optic gyroscope, improve the linearity of the scaling factor, and restrain the zero drift of the open-loop fiber-optic gyroscope, i.e., the dynamic fluctuation of the scaling factor. The fiber-optic gyroscope proposed by the present invention provides a first harmonic demodulation reference signal and a second harmonic demodulation reference signal, which are high in quality and synchronous in detection signal, to the signal processing device proposed by the present invention by the digital phase-locked loop technology.

Abstract: An all-fiber interferometric fiber optic gyroscope having a minimum reciprocal configuration is described. The gyroscope comprises a polarized light source, a light detector, a light source coupler, a fiber optic loop coupler, and a polarization maintaining fiber optic loop. A first port of the light source coupler is counter-axially coupled to an output end of the polarized light source, and a second port of the light source coupler on the same side as the first port is coupled to the light detector. A third port on the other side of the light source coupler is counter-axially coupled to the fiber optic loop coupler, and the fiber optic loop coupler is counter-axially coupled to the polarization maintaining fiber optic loop. The light source splits the input polarized light and polarizes the optical signal propagated along a transmission arm alone, where the first and third ports are on the same transmission arm.

Abstract: An all-fiber interferometric fiber optic gyroscope having a minimum reciprocal configuration is described. The gyroscope comprises a polarized light source, a light detector, a light source coupler, a fiber optic loop coupler, and a polarization maintaining fiber optic loop. A first port of the light source coupler is counter-axially coupled to an output end of the polarized light source, and a second port of the light source coupler on the same side as the first port is coupled to the light detector. A third port on the other side of the light source coupler is counter-axially coupled to the fiber optic loop coupler, and the fiber optic loop coupler is counter-axially coupled to the polarization maintaining fiber optic loop. The light source splits the input polarized light and polarizes the optical signal propagated along a transmission arm alone, where the first and third ports are on the same transmission arm.

Abstract: An all-fiber interferometric fiber optic gyroscope having a minimum reciprocal configuration is described. The gyroscope comprises a light source, a light detector, a light source coupler, a fiber optic loop coupler, and a polarization maintaining fiber optic loop. A first port of the light source coupler is coupled, with polarization axis alignment, to an output end of the light source, and a second port of the light source coupler on the same side as the first port is coupled to the light detector. A third port on the other side of the light source coupler is coupled, with polarization axis alignment, to the fiber optic loop coupler, and the fiber optic loop coupler is coupled, with polarization axis alignment, to the polarization maintaining fiber optic loop. The light source splits the input light and polarizes the optical signal propagated along a transmission arm alone, where the first and third ports are on the same transmission arm.

Abstract: A method for inhibiting zero drift of an all-fiber interferometric fiber optic gyroscope and a corresponding all-fiber interferometric fiber optic gyroscope are disclosed. The method comprises: reversing the polarity of an AC voltage applied to a PZT piezoelectric ceramic phase modulator according to a predetermined half-cycle time period, and making half of the difference between output rotation rates of the gyroscope in two adjacent half-cycle time periods as the output rotation rate of the gyroscope in a cycle. A phase reversal switch and a DSP chip are added to the all-fiber interferometric fiber optic gyroscope. The phase reversal switch is used for controlling the polarity of the AC voltage, and the DSP chip is used for outputting a square wave signal to control the phase reversal switch and for calculating the output rotation rate of the gyroscope according to the output signal of a demodulation/amplifier circuit.