A computer-implemented method for determining an adversarial perturbation for input signals, especially sensor signals or features of sensor signals, of a machine learning system. A best perturbation is determined iteratively, wherein the best perturbation is provided as adversarial perturbation after a predefined amount of iterations, wherein at least one iteration includes: sampling a perturbation; applying the sampled perturbation to an input signal thereby determining a potential adversarial example; determining an output signal from the machine learning system for the potential adversarial example, determining a loss value characterizing a deviation of the output signal to a desired output signal, wherein the desired output signal corresponds to the input signal, if the loss value is larger than a previous loss value setting the best perturbation to the sampled perturbation.

A nonlinear disturbance rejection control apparatus and method for electronic throttle control systems are invented to control the electronic throttle system and to achieve a continuous finite-time disturbance rejection control goal. A control sub-apparatus and method are proposed with an observing sub-apparatus and method for controlling the opening angle of an electronic throttle valve. A mathematical model of the electronic throttle system is analyzed and a control-oriented model is presented with the formation of a lumped disturbance. With combination of the continuous terminal sliding mode control method and the output feedback control method, based on the finite-time high-order sliding mode observer, the preferred control performance is guaranteed, where both the dynamic and static performance of the system is effectively improved.

The present invention is directed towards methods and systems for characterizing sensors and developing classifiers for sensor responses on a utility grid. Experiments are conducted by selectively varying utility grid parameters and observing the responses of utility grid to the variation. Methods and systems of this invention then associate the particular responses of the utility grid sensors with specific variations in the grid parameters, based on knowledge of the areas of space and periods of time where the variation in grid parameters may affect the sensor response. This associated data is then used to updating a model of grid response.

In one embodiment, a network assurance service receives data regarding a monitored network. The service analyzes the received data using a machine learning-based model, to perform a network assurance function for the monitored network. The service detects a lowered performance of the machine learning-based model when a performance metric of the machine learning-based model is below a threshold for the performance metric. When it is determined that the lowered performance of the machine-learning based model is correlated with the sample rate of the received data, the service adjusts the sample rate of the data.

In order to carry out largely automated parameterisation of the closed-loop control parameters for closed-loop control of a hydraulic system comprising a servo drive, a method and a device for determining the closed-loop parameters of a closed-loop control unit of the hydraulic system are specified, wherein an actual system pressure of a hydraulic consumer of the hydraulic system is closed-loop controlled by means of a predefined set point rotational speed of a servo drive, wherein an actual rotational speed of the servo drive follows the predefined set point rotational speed, wherein an excitation signal is applied to the setpoint rotational speed, and the actual system pressure which is set here is measured, the dynamics of the hydraulic system are acquired from the actual rotational speed and/or the setpoint rotational speed and the actual system pressure, and the closed-loop control parameters are calculated from the acquired dynamics.

The techniques disclosed herein enable utilizing a full range of setpoint values to control a mechanical system. A machine learning model is trained with states collected from the mechanical system. Some of the states may have little to no variation, limiting exploration of possible setpoint values when training the model. To enable a more thorough exploration of possible setpoint values, the states are augmented with a fluctuating delta value that is derived from a fixed setpoint value. For example, a delta outside air temperature may be computed by subtracting outside air temperature, which fluctuates, from a fixed chilled water setpoint. A method of moments computation converts delta values inferred by the model back into absolute values. The absolute values are used to compute a regression equation that is usable by the mechanical system to compute a setpoint action for a given set of input states.

Systems and methods for managing an execution of an action strategy by an autonomous system are disclosed. The action strategy comprises a series of actions to be performed by the autonomous system to accomplish a corresponding active objective. The method comprises identifying, by a processor of the autonomous system, an active objective to be accomplished by the autonomous system, the active objective describing a hierarchy of actions to be performed to accomplish the corresponding active objective. The method comprises generating, by the processor, an action strategy from the hierarchy of actions of the active objective, the actions of the action strategy corresponding to the actions of the hierarchy of actions of the active objective and executing the action strategy. Upon completion of an execution of an action, the processor provides data comprising information relating to a state of completion of the action strategy.

A system and a method for generating a navigation path for an autonomous system. The method comprises receiving data comprising characteristics of entities, the entities defining an environment in which the autonomous system is configured to operate; receiving first instructions causing the autonomous system to identify a destination in the environment; generating a navigation path comprising waypoints to be followed by the autonomous system to reach the destination, the waypoints being generated based on the characteristics of the entities and defining segmental paths; executing second instructions causing the autonomous system to navigate along the navigation path; and upon navigating from a first waypoint to a second waypoint: accessing updates of the characteristics of the entities located in a vicinity of a corresponding segmental path; generating a sub-path between the first waypoint and the second waypoint based on second information; and navigating along the sub-path to reach the second waypoint.

Methods of and systems for knowledge-based reasoning to establish a list of active objectives by an autonomous system. The method comprises accessing a list of active objectives; accessing a first database populated with static environment properties, the static environment properties defining properties of entities, the entities defining an environment in which the autonomous system is configured to operate; accessing a second database populated with dynamic environment properties comprising third computer-readable instructions generated by the autonomous system based on events having been observed by the autonomous system. Upon observing a new event, a new dynamic environment property is generated based on the new event and coherence checking is executed to assess whether the new dynamic environment property conflicts with at least one of the static environment properties, and, if so, the new dynamic environment property being identified as incoherent.

A periodic disturbance suppressing control apparatus is designed to estimate and correct an inverse model of a transfer characteristic of a real system successively even in case of large condition change in the real system, and to realize a stable control system. A periodic disturbance of an object to be suppressed is outputted as a sensed periodic disturbance ISdn, ISqn of a direct current component. A difference between a signal obtained by multiplication of the sensed periodic disturbance ISdn, ISqn with a multiplier using a reciprocal Qn of a transfer characteristic, and a signal obtained by adding only a detection delay to a periodic disturbance suppressing command I*dn, I*qn, to estimate the periodic disturbance dI^dn, dI^qn. Thee periodic disturbance suppressing command is calculated by calculating a deviation between the estimated periodic disturbance dI^dn, dI^qn. A learning control section 29 corrects the reciprocal Qn of the transfer characteristic in accordance with a quantity obtained by diving a difference of the periodic disturbance suppressing command I*dn, I*qn during one sample interval by a difference of the sensed periodic disturbance during the one sample interval.