Abstract: A computer-implemented method for failure classification of a surface treatment process includes receiving one or more process parameters that influence one or more failure modes of the surface treatment process and receiving sensor data pertaining to measurement of one or more process states pertaining to the surface treatment process. The method includes processing the received one or more process parameters and the sensor data by a machine learning model deployed on an edge computing device controlling the surface treatment process to generate an output indicating, in real-time, a probability of process failure via the one or more failure modes. The machine learning model is trained on a supervised learning regime based on process data and failure classification labels obtained from physics simulations of the surface treatment process in combination with historical data pertaining to the surface treatment process.

Abstract: A method, a device and a computer program product for abnormal air pollution emission prediction are proposed. In the method, a first set of features characterizing air condition in a zone is obtained. Whether the zone is subject to abnormal air pollution emission in a future first time period is determined based on the first set of features and using a first prediction classifier. In response to determining that the zone is subject to abnormal air pollution emission in the first time period, a second set of features characterizing air condition in the zone is obtained. A future second time period in which the zone is subject to abnormal air pollution emission is determined based on the second set of features and using a second prediction classifier. The second time period is included in the first time period. In this way, the abnormal air pollution emission in the zone can be accurately and efficiently predicted.

Abstract: An edge device can be configured to perform industrial control operations within a production environment that defines a physical location. The edge device can include a plurality of neural network layers that define a deep neural network. The edge device be configured to obtain data from one or more sensors at the physical location defined by the production environment. The edge device can be further configured to perform one or more matrix operations on the data using the plurality of neural network layers so as to generate a large scale matrix computation at the physical location defined by the production environment. In some examples, the edge device can send the large scale matrix computation to a digital twin simulation model associated with the production environment, so as to update the digital twin simulation model in real time.

Abstract: A computer-implemented method includes operating a controllable physical device to perform a task. The method also includes miming forward simulations of the task by a physics engine based on one or more physics parameters. The physics engine communicates with a parameter data layer where each of the one or more physics parameters is modeled with a probability distribution. For each forward simulation run, a tuple of parameter values is sampled from the probability distribution of the one or more physics parameters and fed to the physics engine. The method includes obtaining an observation pertaining to the task from the physical environment and a corresponding forward simulation outcome associated with each sampled tuple of parameter values. The method then includes updating the probability distribution of the one or more physics parameters in the parameter data layer based on the observation from the physical environment and the corresponding forward simulation outcomes.

Abstract: Robots might interact with planar objects (e.g., garments) for process automation, quality control, to perform sewing operations, or the like. It is recognized herein that robots interacting with such planar objects can pose particular problems, for instance problems related to detecting the planar object and estimating the pose of the detected planar object. A system can be configured to detect or segment planar objects, such as garments. The system can include a three-dimensional (3D) sensor positioned to detect a planar object along a transverse direction. The system can further include a first surface that supports the planar object. The first surface can be positioned such that the planar object is disposed between the first surface and the 3D sensor along the transverse direction. In various examples, the 3D sensor is configured to detect the planar object without detecting the first surface.

Abstract: Distributed neural network boosting is performed by a neural network system through operating at least one processor. A method comprises providing a boosting algorithm that distributes a model among a plurality of processing units each being a weak learner of multiple weak learners that can perform computations independent from one another yet process data concurrently. The method further comprises enabling a distributed ensemble learning which enables a programmable logic controller (PLC) to use more than one processing units of the plurality of processing units to scale an application and training the multiple weak learners using the boosting algorithm. The multiple weak learners are machine learning models that do not capture an entire data distribution and are purposefully designed to predict with a lower accuracy. The method further comprises using the multiple weak learners to vote for a final hypothesis based on a feed forward computation of neural networks.

Abstract: A system for supporting artificial intelligence inference in an edge computing device associated with a physical process or plant includes a neural network training module, a neural network testing module and a digital twin of the physical process or plant. The neural network training module is configured to train a neural network model for deployment to the edge computing device based on data including baseline training data and field data received from the edge computing device. The neural network testing module configured to validate the trained neural network model prior to deployment to the edge computing device by leveraging the digital twin of the physical process or plant.

Abstract: According to an aspect of the present disclosure, a computer-implemented includes creating a plurality of basic skill functions for a controllable physical device of an autonomous system. Each basic skill function includes a functional description for using the controllable physical device to interact with a physical environment to perform a defined objective. The method further includes selecting one or more basic skill functions to configure the controllable physical device to perform a defined task. The method also includes determining a decorator skill function specifying at least one constraint. The decorator skill function is configured to impose, at run-time, the at least one constraint, on the one or more basic skill functions. The method further includes generating executable code by applying the decorator skill function to the one or more basic skill functions, and actuating the controllable physical device using the executable code.

Abstract: In current applications of autonomous machines in industrial settings, the environment, in particular the devices and systems with which the machine interacts, is known such that the autonomous machine can operate in the particular environment successfully. Thus, current approaches to automating tasks within varying environments, for instance complex environments having uncertainties, lack capabilities and efficiencies. In an example aspect, a method for operating an autonomous machine within a physical environment includes detecting an object within the physical environment. The autonomous machine can determine and perform a principle of operation associated with a detected subcomponent of the object, so as to complete a task that requires that the autonomous machine interacts with the object. In some cases, the autonomous machine has not previously encountered the object.

Abstract: A method for automatically generating a bill of process in a manufacturing system comprising: receiving design information representative of a product to be produced; iteratively performing simulations of the manufacturing system; identifying manufacturing actions based on the simulations; optimizing the identified manufacturing actions to efficiently produce the product to be produced; generating, by the manufacturing system, a bill of process for producing the product. Simulations may be performed using a digital twin of the product being produced and a digital twin of the environment. System actions are optimized using a reinforcement learning technique to automatically produce a bill of process based on the design information of the product and task specifications.

Abstract: A computer-implemented method for designing execution of a process by a robotic cell includes obtaining a process goal and one or more process constraints. The method includes accessing a library of constructs and a library of skills. Each construct includes a digital representation of a component of the robotic cell or a geometric transformation of the robotic cell. Each skill includes a functional description for using a robot of the robotic cell to interact with a physical environment to perform a skill objective. The method uses a simulation engine to simulate a multiplicity of designs, wherein each design is characterized by a combination of constructs and skills to achieve the process goal, and determine a set of feasible designs that meet the one or more process constraints. The method includes outputting recommended designs from the set of feasible designs.

Abstract: Multi-level objectives improve efficiency of multi-objective automated machine learning. A hyperband framework is established with a kernel density estimator to shrink the search space based on evaluation of lower-level objectives. A Gaussian prior assumption directly shrinks the search space to find a main objective.

Abstract: A method for optimizing a neural network architecture by estimating an inference time for each operator in the neural network architecture is provided. The method may include determining a benchmark time for at least one single-path architecture out of a plurality of single-path architectures associated with the neural network by sampling the at least one single-path architecture from the neural network, wherein the at least one single-path architecture comprises one or more operators. The method may further include, based on the benchmark time for the at least one single-path architecture, determining an estimated inference time for an operator, wherein determining the estimated inference time for the operator comprises, applying an operator function, wherein the operator function comprises a function based on a difference between the benchmark time associated with the at least one single-path architecture and the estimated latency of the neural network.

Abstract: An abnormal area is detected using an initial spatial weights matrix between pairs of air quality sensors in a plurality of air quality sensors distributed across a geographical area and air quality data for each air quality sensor. The spatial weights matrix utilizes a distance between pairs of air quality sensors and wind direction through the geographical area. The initial spatial weights matrix and air quality data are used to calculate a plurality of local moran's indexes, one for each air quality sensor. The plurality of local moran's indexes are used to divide the plurality of air quality sensors into four groups. The groups are classified as proper or improper, and the proper groups are identified as abnormal areas.

Abstract: Correction management techniques are provided. In one embodiment, the techniques involve determining, via a first machine learning model, a first and second data based on a respective first and second raw data obtained from a plurality of sensors, determining, based on a deviation between the first data and the second data, an inaccuracy of the first data, identifying an ambient situation corresponding to the first raw data and the second raw data, selecting, from historical raw data of the plurality of sensors, a group of raw data corresponding to the ambient situation, and correcting, via a second machine learning model, the first data based on the selected group of raw data.

Abstract: Systems and methods provide real-time production scheduling by integrating deep reinforcement learning and Monte Carlo tree search. A manufacturing process simulator is used to train a deep reinforcement learning agent to identify the sub-optimal policies for a production schedule. A Monte Carlo tree search agent is implemented to speed up the search for near-optimal policies of higher quality from the sub-optimal policies.

Abstract: An abnormal area is detected using an initial spatial weights matrix between pairs of air quality sensors in a plurality of air quality sensors distributed across a geographical area and air quality data for each air quality sensor. The spatial weights matrix utilizes a distance between pairs of air quality sensors and wind direction through the geographical area. The initial spatial weights matrix and air quality data are used to calculate a plurality of local moran's indexes, one for each air quality sensor. The plurality of local moran's indexes are used to divide the plurality of air quality sensors into four groups. The groups are classified as proper or improper, and the proper groups are identified as abnormal areas.

Abstract: A method, a device and a computer program product for abnormal air pollution emission prediction are proposed. In the method, a first set of features characterizing air condition in a zone is obtained. Whether the zone is subject to abnormal air pollution emission in a future first time period is determined based on the first set of features and using a first prediction classifier. In response to determining that the zone is subject to abnormal air pollution emission in the first time period, a second set of features characterizing air condition in the zone is obtained. A future second time period in which the zone is subject to abnormal air pollution emission is determined based on the second set of features and using a second prediction classifier. The second time period is included in the first time period. In this way, the abnormal air pollution emission in the zone can be accurately and efficiently predicted.

Abstract: Embodiments of the present invention relate to methods, systems, and computer program products for correction management. In a method, a deviation may be detected by one or more processors between a first data obtained from a sensor in a plurality of sensors and a second data obtained from other sensors in the plurality of sensors, the first data and the second data being obtained in an identical or similar ambient situation. The ambient situation where the first data and second data are obtained may be identified by one or more processors. A group of raw data that is monitored under the ambient situation may be selected by one or more processors from historical raw data monitored by the plurality of sensors. The first data may be corrected by one or more processors based on the selected group of raw data.

Abstract: In an embodiment of the present disclosure, a method for modeling prediction of inhalable particles concentration is disclosed. In the method, at least one dispersal event is identified, and at least one accumulation event is identified based on the identified at least one dispersal event. Then a dispersal prediction model is generated based on the identified at least one dispersal event. Then at least one accumulation level of inhalable particles concentration is obtained for the at least one accumulation event. A change prediction model for the accumulation level is generated. Then a plurality of accumulation prediction models is generated.