Abstract: Apparatuses and methods are disclosed for dynamic data package routing. A method includes receiving criterion dynamics, constructing Dynkin operators in response to the criterion dynamics, constructing an interference matrix in response to the Dynkin operators, deploying an interference automaton equation in response to the interference matrix, performing prefix-loop decomposition to generate a control law, and outputting the control law.

Abstract: Techniques are described for implementing an automated control system to control operations of a target physical system, such as production of electrical power in an electrical grid. The techniques may include determining how much electrical power for each of multiple producers to supply for each of a series of time periods, such as to satisfy projected demand for that time period while maximizing one or more indicated goals, and initiating corresponding control actions. The techniques may further include repeatedly performing automated modifications to the control system's ongoing operations to improve the target system's functionality, by using reinforcement learning to iteratively optimize particles generated for a time period that represent different state information within the target system, to learn one or more possible solutions for satisfying projected electrical power load during that time period while best meeting the one or more defined goals.

Abstract: Techniques are described for implementing automated control systems to control operations of target physical systems and/or their components (e.g., a fuel cell, wind turbine, HVAC unit, etc.), such as based at least in part on models of their dynamic non-linear behaviors that are generated by gathering and analyzing information about their operations under varying conditions. The techniques may include, for each of multiple levels of inputs to the system/component and/or other factors, injecting a corresponding signal input into the system/component, and using active sensors to collect time changes of the responses to these pulses. Information about the inputs and the responses is used to generate an incremental parametric model representing the internal state and behavioral dynamics of the system/component, which is further used to control additional ongoing operations of the system/component (e.g., to control whether and how much output is produced in a current or future time period).

Abstract: Techniques are described for implementing an automated control system to control operations of a target physical system, such as production of electrical power in an electrical grid. The techniques may include determining how much electrical power for each of multiple producers to supply for each of a series of time periods, such as to satisfy projected demand for that time period while maximizing one or more indicated goals, and initiating corresponding control actions. The techniques may further include repeatedly performing automated modifications to the control system's ongoing operations to improve the target system's functionality, by using reinforcement learning to iteratively optimize particles generated for a time period that represent different state information within the target system, to learn one or more possible solutions for satisfying projected electrical power load during that time period while best meeting the one or more defined goals.

Abstract: Techniques are described for implementing automated control systems that manipulate operations of specified target systems, such as by modifying or otherwise manipulating inputs or other control elements of the target system that affect its operation (e.g., affect output of the target system). An automated control system may in some situations have a distributed architecture with multiple decision modules that each controls a portion of a target system and operate in a partially decoupled manner with respect to each other, such as by each decision module operating to synchronize its local solutions and proposed control actions with those of one or more other decision modules, in order to determine a consensus with those other decision modules. Such inter-module synchronizations may occur repeatedly to determine one or more control actions for each decision module at a particular time, as well as to be repeated over multiple times for ongoing control.

Abstract: Techniques are described for implementing automated control systems that each control or otherwise manipulate, for a target system having one or more battery cells each having internal components that include one or more internal supercapacitor components in parallel with at least one battery component, usage operations for one of the internal components of one of the battery cells, with the usage operations for the internal components of a particular battery cell being synchronized or otherwise coordinated to protect the battery component(s) of the battery cell while satisfying other criteria (e.g., to increase battery cell life and/or reduce power dissipation). In at least some situations, the target system is an electric vehicle, and the automated control systems control the electric vehicle's battery cells to provide electrical power to the motor during acceleration and constant speed driving, and to store electrical power in the battery cells during braking or other deceleration.

Abstract: Techniques are described for implementing automated control systems to control operations of target physical systems including batteries, such as based at least in part on models of the batteries' dynamic behaviors that are generated by gathering and analyzing information about the batteries' operations under varying conditions. The techniques may include, for each of multiple charge levels of a battery and for each of multiple resistive loads, injecting multi-frequency microvolt pulses into the battery, and using sensors to collect time changes of the responses to these pulses. Information about the inputs and the responses is then analyzed and used to generate an incremental parametric state model representing the internal dynamics of the battery, which is further used to control additional ongoing battery operations (e.g., to control whether and how much power is supplied to and/or extracted from the battery in a current or future time period).

Abstract: An automated control system to control at least some operations of one or more target physical systems that each includes one or more batteries. The described techniques may include determining whether and how much power to supply for each of a series of time periods, and implementing the determined power amount for a time period by determining and setting one of multiple impedance level control values of an associated actuator component. Repeated automated operations of this type may include using parametric linear approximation to determine one of multiple enumerated control values that best satisfies one or more defined goals at a given time in light of current state information (e.g., current output from the battery, voltage from the battery, battery temperature, etc.), such as by repeatedly determining an improved distribution function over the control values, and propagating it over multiple future time periods.

Abstract: Techniques are described for implementing automated control systems that manipulate operations of specified target systems, such as by modifying or otherwise manipulating inputs or other control elements of the target system that affect its operation (e.g., affect output of the target system). An automated control system may in some situations have a distributed architecture with multiple decision modules that each controls a portion of a target system, and may further have one or more components that interacts with one or more users to obtain a description of the target system, including restrictions related to the various elements of the target system, and one or more goals to be achieved during control of the target system. The component(s) then perform various automated actions to generate, test and deploy one or more executable decision modules to use in performing the control of the target system based on the user-specified information.

Abstract: Techniques are described for implementing automated control systems for target battery systems based at least in part on battery state information gathered from active excitation of the batteries, such as to maximize battery life while performing other battery power use activities. The excitation of a target battery system may occur while it is in use, by repeatedly introducing small defined variations as input to the battery system while the battery system is otherwise used to supply or receive electricity.

Abstract: Techniques are described for implementing automated control systems that repeatedly perform automated modifications to control system actuator components' ongoing operations to improve functionality for target battery systems, such as to reduce power dissipation while performing other battery power use activities to maximize battery life. Controlling a battery's usage may include using a DC-to-DC amplifier, and the repeated automated modifications may include modifying the state of the DC-to-DC amplifier actuator to adjust a level of resistance and/or an amount of time during which power is supplied. The repeated automated modifications may be performed to repeatedly reduce the distance between the current battery performance and an idealized version of the battery performance (e.g., a version with no power dissipation).

Abstract: Techniques are described for implementing automated control systems that repeatedly perform automated modifications to control system actuator components' ongoing operations to improve functionality for electrical devices in target systems, such as to reduce power dissipation while using the electrical devices. The described techniques further include synchronizing a particular control system's state improvements with corresponding control system state improvements being performed for one or more other control systems that are each controlling one or more distinct electrical devices in the target system(s), so as to improve the collective control system functionality according to one or more criteria (e.g., to reduce power dissipation)—such synchronizing may include, for example, generating a mean field representation of the overall control system state for the target system(s) and using the mean field representation to improve the overall control system state.

Abstract: Methods and systems for transcribing a media file using a human intelligence task service and/or reinforcement learning are provided. The disclosed systems and methods provide opportunities for a segment of the input media file to be automatically re-analyzed, re-transcribed, and/or modified for re-transcription using a human intelligence task (HIT) service for verification and/or modification of the transcription results. The segment can also be reanalyzed, reconstructed, and re-transcribed using a reinforcement learning enabled transcription model.

Abstract: Techniques are described for implementing automated control systems for target battery systems based at least in part on battery state information gathered from active excitation of the batteries, such as to maximize battery life while performing other battery power use activities. The excitation of a target battery system may occur while it is in use, by repeatedly introducing small defined variations as input to the battery system while the battery system is otherwise used to supply or receive electricity. Corresponding small variations in output of the battery system from the excitation activities are then measured by hardware sensors, aggregated and analyzed to generate a current model of the internal state of the one or more batteries, and then used to assist in controlling further operations of the battery system, including in some cases to update a previously existing model of the battery system.

Abstract: Techniques are described for implementing automated control systems to control operations of specified physical target systems, such as with one or more batteries used to store and provide electrical power. Characteristics of each battery's state may be used to perform automated control of DC power from the battery, such as in a real-time manner and to optimize long-term operation of the battery. In some situations, multiple batteries are controlled by using multiple control systems each associated with one of the batteries, and with overall control being coordinated in a distributed manner using interactions between the multiple control systems. A system that includes one or more batteries to be controlled may further include additional components in some situations, such as one or more electrical sources and/or one or more electrical loads, with one non-exclusive example of a type of such system being one or more home electrical power systems.

Abstract: Techniques are described for implementing automated control systems to control operations of specified physical target systems. In some situations, the described techniques include forecasting future values of parameters that affect operation of a target system, and using the forecasted future values as part of determining current automated control actions to take for the target system—in this manner, the current automated control actions may be improved relative to other possible actions that do not reflect such forecasted future values. Various automated operations may also be performed to improve the forecasting in at least some situations, such as by combining the use of multiple different types of forecasting models and multiple different groups of past data to use for training the models, and/or by improving the estimated internal non-observable state information reflected in at least some of the models.

Abstract: Methods and systems for transcribing a media file using reinforcement learning are provided. In one aspect, the method includes: identifying a low confidence of accuracy portion from a transcription result of the media file; constructing a phoneme sequence that includes an audio segment corresponding to the identified low confidence of accuracy portion, based on at least on a reward function; creating a new audio waveform based at least on the constructed phoneme sequence; and generating a new transcription using a transcription engine based on the new audio waveform.

Abstract: Techniques are described for implementing automated control systems that repeatedly perform automated modifications to control system actuator components' ongoing operations to improve functionality for target battery systems, such as to reduce power dissipation while performing other battery power use activities to maximize battery life. Controlling a battery's usage may include using a DC-to-DC amplifier, and the repeated automated modifications may include modifying the state of the DC-to-DC amplifier actuator to adjust a level of resistance and/or an amount of time during which power is supplied. The repeated automated modifications may be performed to repeatedly reduce the distance between the current battery performance and an idealized version of the battery performance (e.g., a version with no power dissipation).

Abstract: Techniques are described for implementing automated control systems to control operations of specified physical target systems. In some situations, the described techniques include obtaining and analyzing sensor data about operations of a target system in order to generate an improved model of a current state of the target system, and using the modeled state information as part of determining further current and/or future automated control actions to take for the target system, such as to generate a function and/or other structure that models internal operations of the target system, rather than merely attempting to estimate target system output from input without understanding the internal structure and operations of the target system.

Abstract: A set of SKUs is divided into a plurality of different Mean Field clusters, and a tracker (or sensor) is identified for each cluster. Product decisions for each Mean Field cluster are generated based on the tracker (or sensor) and each Mean Field cluster is then deconstructed to obtain product decisions for individual SKUs in the Mean Field cluster.