Smart Systems ME 4SS3 Course Overview

Smart systems have five key components: Perception, Control, Knowledge, Communication, and [missing component]

The "smartness" of a system depends on its level of autonomous operation related to the application, and how well it can react and predict outcomes.

The ability of a smart system to sense and gather information about its environment.

The ability of a smart system to make decisions and adjust actions based on gathered information to control its environment and achieve goals.

The information and rules a smart system uses to make decisions and act effectively. This includes the data and data-processing elements.

Describes how a signal's amplitude changes with respect to its frequency.

Describes how a signal changes over time.

A set of first-order differential equations describing a system's behavior using states (variables like position, velocity).

A way of mathematically representing a physical system with states, state vectors, and output equations.

The frequency at which a system oscillates without any damping forces.

The ability to determine the state of a dynamic system uniquely from its inputs and outputs, based on a known system model.

A system whose state cannot be uniquely determined from its inputs and outputs, even with a known model.

A mathematical tool used to determine whether a system is observable or not, based on its model.

The process of extracting true state knowledge from noisy or corrupted signals.

Unwanted variations or disturbances that affect the system's state, like a rumble in a transmission.

Noise introduced by the measurement process itself, like faulty sensors causing inaccurate readings.

A type of noise with a constant power spectral density across all frequencies, like static on a television.

An equation that estimates the system's state using known inputs and model parameters, even with noise.

The Kalman filter estimates the state of a system (like position, velocity) by combining noisy sensor measurements with a model of the system's dynamics. It finds the best balance between the two to produce the most accurate state estimate.

The Kalman gain determines how much weight is given to new sensor measurements in updating the state estimate. A higher Kalman gain means more trust is given to the new measurements, while a lower gain means more reliance on the system model.

The state error covariance matrix represents the uncertainty in the state estimate. It tells us how much the estimated state could be off from the true state.

The prediction step uses the system model to predict the state of the system at the next time step, based on the current state estimate.

The update step corrects the predicted state by incorporating the new sensor measurements. It combines the predicted state with the new measurements to get an updated estimate.

Machine learning is a type of artificial intelligence where algorithms are trained on datasets to learn patterns and make predictions or decisions.

Deep learning is a subset of machine learning that uses artificial neural networks with multiple layers to learn complex patterns in data.

A system that integrates sensing, control, communication, and knowledge to adapt and interact with its environment autonomously.

The process of creating a mathematical representation of a physical system's behavior, often using equations and diagrams.

A powerful algorithm used to estimate the true state of a system, especially in the presence of noise, by combining model predictions with sensor measurements.

The slope of the error curve represents the rate of change of error over time. It's calculated as the difference in error between two consecutive time steps divided by the time interval.

The derivative of the error term is the instantaneous rate of change of error. It tells you how fast the error is changing at a particular point in time.

The integral of the error term is the area under the error curve. It represents the accumulated error over time.

A discrete-time PID controller modifies a system's output based on the current error, the derivative of the error, and the integral of the error, in order to reach a desired setpoint.

The proportional term in a PID controller amplifies the current error by a constant factor (Kp). Larger Kp values result in a faster response.

The derivative term in a PID controller uses the rate of change of error to predict future error and provide a stabilizing influence.

The integral term in a PID controller accumulates the error over time, providing a corrective action to compensate for any steady-state error.

To extract true state knowledge from noisy or corrupted signals in a system.

A statistical algorithm that estimates the true state of a system by combining noisy sensor measurements with a model of the system's dynamics. It yields the optimal solution based on Gaussian noise and known system parameters.

An estimator that minimizes the mean-squared error between the estimated and true state values. It aims to find the best estimate on average.

The Kalman filter relies on several key assumptions: known system parameters, Gaussian and white noises, and exact knowledge of covariances. These assumptions are necessary for optimal performance.

A vector representing the current state of a system. It contains variables like position, velocity, or other relevant characteristics.

A matrix representing the uncertainty in the state vector. It indicates how much the estimated state could deviate from the true state.

A parameter used to adjust the balance between the predicted state and the new measurements. A higher gain gives more weight to new data, while a lower gain relies more on the model.

The step where the Kalman filter estimates the state at the next time step based on the previous state and the system model.

The step where the Kalman filter corrects the predicted state by incorporating new sensor measurements.

The problem of estimating the state of a linear system based on noisy observations. The Kalman filter is the optimal solution for this problem.