ME 4SS3 L27 Smart Systems Review PDF

Smart Systems ME 4SS3 Dr. S. Andrew Gadsden Department of Mechanical Engineering, McMaster University L27.1 Review of Smart Systems Material L27.2 Midterm Reminders L27.3 Course Wrap-up  Monday...

Smart Systems ME 4SS3 Dr. S. Andrew Gadsden Department of Mechanical Engineering, McMaster University L27.1 Review of Smart Systems Material L27.2 Midterm Reminders L27.3 Course Wrap-up  Monday Wednesday Thursday Deliverables (Virtual on A2L or MS Teams) (HH 305) (HH 305) - - L01: Introduction 09/04 L02: Introduction to 09/05 - to Course Project L03: Introduction to 09/09 L04: System 09/11 L05: Project - Laser 09/12 - Smart Systems Modeling I Cutting (JHE A104A) L06: System Modeling 09/16 L07: System Modeling 09/18 L08: Project - Assembly 09/19 - Practice II Check and Testing L09: System Modeling 09/23 L10: System Modeling 09/25 L11: Project - Modeling 09/26 Assignment 1 and Matlab I and Matlab II Examples (09/29) No Class 09/30 L12: Control Theory 10/02 L13: Project - Matlab 10/03 - (Truth and Simulation Reconciliation) L14: Signal 10/07 L15: Controllers 10/09 L16: Project - 10/10 - Conditioning and Matlab Arduino Tutorial Where are we? No Class 10/14 No Class 10/16 No Class 10/17 Assignment 2 (Thanksgiving Monday) (Break) (Break) (10/20) L17: Introduction to 10/21 L18: Kalman Filter I 10/23 L19: Project – TA 10/24 - Estimation Theory Consultations L20: Kalman Filter II 10/28 L21: Kalman Filter 10/30 L22: Project - TA 10/31 Assignment 3 L27 - Smart Systems and Matlab Consultations (11/03) L23: Introduction to 11/04 L24: Resume 11/06 L25: HR Industry 11/07 Assignment 4 Machine Learning Workshop (Online) Panel (In-Person) (11/10) L26: Machine Learning 11/11 L27: Review of Smart 11/13 L28: Project - TA 11/14 Resume Asgmt. Applications Systems Material Consultations (11/17) Virtual Office Hours on 11/18 L29: Midterm Review 11/20 L30: Midterm 11/21 Midterm MS Teams and Help (11/21) Virtual Office Hours on 11/25 L31: Project - TA 11/27 L32: Project - 11/28 Project Demo MS Teams Consultations (11/28) 2 Demonstration Day Virtual Office Hours on 12/02 Virtual Office Hours on 12/04 Project - Report Due 12/05 Project Report MS Teams MS Teams (No Class) (12/05) 27.1 Course Review Smart Systems Smart systems typically incorporate perception and control in order to interact with the environment, and make decisions based on available data in a predictive or adaptive manner  ‘Smartness’ is usually associated with the level of autonomous operations Often, it is dependent on the field or area of application L27 - Smart Systems E.g., smart manufacturing or industry 4.0 3 27.1 Course Review Smart Systems The five main components of a smart system: 1. Perception 2. Control 3. Knowledge 4. Communication L27 - Smart Systems 5. Security The first three components were covered in this course 4 27.1 Course Review Smart Systems Systems may be studied in the frequency or time domain Frequency-domain: how the amplitude of the signal changes with respect to frequency  Transfer functions (Chapter 2 of Nise) Time-domain: how a signal changes over time  State equations (Chapter 3 of Nise) L27 - Smart Systems Equations (either domain) may be used to describe the relationship between the input and output of a system 5 27.1 Course Review Smart Systems An alternate approach to frequency-domain [Chapter 2 of Nise] techniques is modeling systems in the time-domain [Chapter 3 of Nise] State-space representation is a way to mathematically represent a physical system  States: linearly independent system variables (e.g., position, velocity, acceleration) L27 - Smart Systems  State vector: a vector whose elements are states  State equations: a set of n first-order differential equations with n variables (or states)  Output equation: algebraic representation that expresses the output as a combination of states and inputs 6 27.1 Course Review Smart Systems A linear system may be represented in state space by the following equations: 𝒙𝒙̇ = 𝐴𝐴𝒙𝒙 + 𝐵𝐵𝒖𝒖 𝒚𝒚 = 𝐶𝐶𝒙𝒙 + 𝐷𝐷𝒖𝒖 L27 - Smart Systems 𝑥𝑥̇ 1 0 1 0 0 𝑥𝑥1 0 𝑣𝑣̇ 1 −𝐾𝐾/𝑀𝑀1 −𝐷𝐷/𝑀𝑀1 𝐾𝐾/𝑀𝑀1 0 𝑣𝑣1 0 = 𝑥𝑥2 + 𝑓𝑓 𝑡𝑡 𝑥𝑥̇ 2 0 0 0 1 0 𝑣𝑣̇ 2 𝐾𝐾/𝑀𝑀2 0 −𝐾𝐾/𝑀𝑀2 0 𝑣𝑣2 1/𝑀𝑀2 7 27.1 Course Review Smart Systems Two physically meaningful specifications:  Natural frequency, 𝜔𝜔𝑛𝑛  Damping ratio, 𝜁𝜁 Natural frequency: defined as the frequency of oscillation of the system without damping Damping ratio: a dimensionless measure describing how oscillations in a L27 - Smart Systems system decay after a disturbance 8 27.1 Course Review Smart Systems Performance specifications for second-order systems: 1. Natural frequency, 𝜔𝜔𝑛𝑛 2. Damping ratio, 𝜁𝜁 3. Rise time, 𝑇𝑇𝑟𝑟 4. Peak time, 𝑇𝑇𝑝𝑝 : time required to reach the first, or maximum peak L27 - Smart Systems (in terms of output magnitude) 5. Percent overshoot, %𝑂𝑂𝑂𝑂: amount that the waveform overshoots the steady-state, or final, value at the peak time (generally expressed as a percentage of the steady-state) 6. Settling time, 𝑇𝑇𝑠𝑠 9 27.1 Course Review Smart Systems Dynamics are represented by a set of first order difference equations in matrix form: 𝑥𝑥𝑘𝑘+1 = 𝐴𝐴𝑘𝑘 𝑥𝑥𝑘𝑘 + 𝐵𝐵𝑘𝑘 𝑢𝑢𝑘𝑘 𝑧𝑧𝑘𝑘+1 = 𝐶𝐶𝑘𝑘+1 𝑥𝑥𝑘𝑘+1 We can discretize continuous state space models using computer software (e.g., MATLAB), but we should understand the basic process L27 - Smart Systems Important: Not all systems can be represented by differential equations or state-space models  For this course, we will consider those that can be represented (in fact, most can be represented accordingly) 10 27.1 Course Review Smart Systems Discrete-time PID controller: 𝑢𝑢𝑘𝑘 = 𝐾𝐾𝑝𝑝 𝑒𝑒𝑘𝑘 + 𝐾𝐾𝑑𝑑 (𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑜𝑜𝑜𝑜 𝑒𝑒𝑘𝑘 ) + 𝐾𝐾𝑖𝑖 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑜𝑜𝑜𝑜 𝑒𝑒𝑘𝑘 What are the derivative and integral of the error term?  Derivative? Slope of the error curve (rise over run)… 𝑒𝑒 L27 - Smart Systems 𝑒𝑒𝑘𝑘+1 − 𝑒𝑒𝑘𝑘 𝑡𝑡𝑘𝑘 𝑇𝑇 𝑡𝑡𝑘𝑘+1 Error 𝒅𝒅 𝒆𝒆 𝒆𝒆𝒌𝒌+𝟏𝟏 − 𝒆𝒆𝒌𝒌 𝒆𝒆𝒌𝒌+𝟏𝟏 − 𝒆𝒆𝒌𝒌 11 ≅ = 𝒅𝒅𝒅𝒅 𝒕𝒕𝒌𝒌+𝟏𝟏 − 𝒕𝒕𝒌𝒌 𝑻𝑻 Time 27.1 Course Review Smart Systems Discrete-time PID controller: 𝑢𝑢𝑘𝑘 = 𝐾𝐾𝑝𝑝 𝑒𝑒𝑘𝑘 + 𝐾𝐾𝑑𝑑 (𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑜𝑜𝑜𝑜 𝑒𝑒𝑘𝑘 ) + 𝐾𝐾𝑖𝑖 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑜𝑜𝑜𝑜 𝑒𝑒𝑘𝑘 What are the derivative and integral of the error term?  Integral? Area under the error curve…  We should add up this integral error throughout the entire 𝑒𝑒 𝑒𝑒 ≅ 1 + 2 simulation (still an approximation) L27 - Smart Systems 1 = 0.5 𝑒𝑒𝑘𝑘+1 − 𝑒𝑒𝑘𝑘 𝑇𝑇 𝑒𝑒𝑘𝑘+1 − 𝑒𝑒𝑘𝑘 (1) (2) = 𝑒𝑒𝑘𝑘 𝑇𝑇 Error 𝑒𝑒 ≅ 𝑒𝑒𝑘𝑘 𝑇𝑇 + 0.5 𝑒𝑒𝑘𝑘+1 − 𝑒𝑒𝑘𝑘 𝑇𝑇 (2) 𝑒𝑒𝑘𝑘 12 𝒆𝒆 ≅ 𝟎𝟎. 𝟓𝟓 𝒆𝒆𝒌𝒌+𝟏𝟏 + 𝒆𝒆𝒌𝒌 𝑻𝑻 Time 𝑡𝑡𝑘𝑘 𝑇𝑇 𝑡𝑡𝑘𝑘+1 27.1 Course Review Smart Systems Discrete-time PID controller: 𝑒𝑒𝑘𝑘 − 𝑒𝑒𝑘𝑘−1 𝑢𝑢𝑘𝑘 = 𝐾𝐾𝑝𝑝 𝑒𝑒𝑘𝑘 + 𝐾𝐾𝑑𝑑 + 𝐾𝐾𝑖𝑖 𝑒𝑒𝑖𝑖𝑖𝑖𝑖𝑖 𝑇𝑇 Where 𝑒𝑒𝑖𝑖𝑖𝑖𝑖𝑖 = ∑𝑡𝑡𝑘𝑘=1 0.5 𝑒𝑒𝑘𝑘 + 𝑒𝑒𝑘𝑘−1 𝑇𝑇 and is summed throughout time What is the effect of 𝑇𝑇 on the control signal? L27 - Smart Systems 13 27.1 Course Review Smart Systems In general, but not always:  Increasing 𝐾𝐾𝑝𝑝 makes the response faster, and reduces the steady-state error (a bit)  Increasing 𝐾𝐾𝑑𝑑 reduces the overshoot  Increasing 𝐾𝐾𝑖𝑖 reduces the steady-state error Since there is interaction between the three gains, tuning them manually can be very challenging L27 - Smart Systems Engineering design is a skill utilized by your knowledge base! 14 27.1 Course Review Smart Systems The output signal from the sensor of a measurement system has generally to be processed in some way to make it suitable for next stage of operation For example:  The signal may be too small and have to be amplified  Signal may contain interference which needs to be removed  Signal may be nonlinear, and require linearization L27 - Smart Systems  Signal could be analogue, and needs to be made digital  Be a resistance change and have to be made into a current change… All of these scenarios can be referred to as signal conditioning 15 27.1 Course Review Smart Systems The basis of many signal conditioning modules is the operational amplifier The ‘op amp’ is a high-gain D.C. amplifier (gain is typically 100,000 or more) that is supplied as an integrated circuit on a silicon chip It has two inputs:  Inverting input (–) L27 - Smart Systems  Non-inverting input (+) Other inputs may include a positive and negative voltage supply 16 27.1 Course Review Smart Systems The term filtering is used to describe the process of removing a certain band of frequencies from a signal and permitting others to be transmitted The range of frequencies by a filter is known as the pass band, the range not passed as the stop band, and the boundary between stopping and passing as the cut-off frequency L27 - Smart Systems Filters may be classified according to the frequency ranges they transmit or reject  Low-pass, high-pass  Band-pass, band-stop 17 27.1 Course Review Smart Systems Pulse width modulation (PWM) is widely used with control systems as a means of controlling the average value of a D.C. voltage If there is a constant analogue voltage and it is chopped into pulses, the average value of the voltage can be changed by varying the width of the pulses L27 - Smart Systems Duty cycle is the fraction of each cycle for which the voltage is ‘high’ 18 27.1 Course Review Smart Systems Electromagnetic interference is an undesirable effect on circuits resulting from time-varying electric and magnetic fields Common sources include fluorescent lamps, D.C. motors, relay coils, household appliances, and the electrics of ‘motor cars’ Electrostatic interference occurs as a result of mutual capacitance between neighbouring conductors L27 - Smart Systems Interference also occurs when there is a changing magnetic field which induces voltages in the measurement system 19 27.1 Course Review Smart Systems Observability is the issue of whether the state of a dynamic system with a known model is uniquely determinable from its inputs and outputs It is essentially a property of the given system model If the system state is not uniquely determinable from the system inputs and outputs, then the system model is considered L27 - Smart Systems unobservable An observability matrix of the given system model is calculated to determine whether or not a system is completely (or not at all) observable 20 27.1 Course Review Smart Systems If a system is observable, we can utilize estimation theory to gain knowledge of the system states Most (if not all) systems and measurements contain some noise  We do not know their values exactly  Sometimes even the act of measuring a system can change the outcome of the measurement (e.g., subatomic level)  Signal conditioning methods can be used as well L27 - Smart Systems We can use the following nomenclature to model noise: 𝑤𝑤𝑘𝑘 : system noise at time 𝑘𝑘 𝑣𝑣𝑘𝑘 : measurement noise at time 𝑘𝑘 21 27.1 Course Review Smart Systems It is the primary goal of estimation theory to extract true state knowledge from noisy or corrupted signals  E.g., using a radar to track the position of an aircraft  E.g., using an accelerometer to calculate the angular rotation of a beam Improved perception will lead to a better response from the smart system in general  E.g., ‘maximize information’ for improved system performance L27 - Smart Systems 22 27.1 Course Review Smart Systems We can write our linear measurement equation with noise as: 𝑧𝑧𝑘𝑘+1 = 𝐶𝐶𝑥𝑥𝑘𝑘+1 + 𝑣𝑣𝑘𝑘+1 Where 𝑣𝑣𝑘𝑘+1 is the amount of noise in our measurement due to the quality of the sensor itself Typically, we model measurement noise as white noise which means L27 - Smart Systems its distribution is zero mean and has a Gaussian distribution 𝑃𝑃 𝑣𝑣 ~𝒩𝒩 0, 𝑅𝑅 23 27.1 Course Review Smart Systems So, it is the ultimate goal of any estimation strategy to use noise- corrupted measurements 𝑧𝑧 to obtain estimates of our states 𝒙𝒙 The estimated state equation for linear systems is defined as: 𝑥𝑥 𝑘𝑘+1 = 𝐴𝐴𝑥𝑥 𝑘𝑘 + 𝐵𝐵𝑢𝑢𝑘𝑘 And, for completeness, our estimated measurement equation becomes: L27 - Smart Systems 𝑧𝑧̂𝑘𝑘+1 = 𝐶𝐶 𝑥𝑥 𝑘𝑘+1 Note: 𝑘𝑘 is time step, and the above assumes we know 𝐴𝐴, 𝐵𝐵, and 𝐶𝐶 Since we have no actual knowledge of 𝑤𝑤 and 𝑣𝑣, we can use statistics (or even machine learning) to approximate 𝒙𝒙 from 𝒛𝒛  Normal distributions require knowledge of the mean and variance! 24 27.1 Course Review Smart Systems The Kalman filter (KF) yields the optimal stochastic solution to the linear estimation problem  It is a minimum mean-square error (MMSE) estimator It yields the optimal solution based on a few assumptions: 1. The system, gain, and measurement matrices are known (i.e., there is no modeling uncertainty) 2. The system and measurement noises are Gaussian, white L27 - Smart Systems 3. The system and measurement covariances are exactly known 25 27.1 Course Review Smart Systems The probability distribution of the state vector 𝒙𝒙 can be effectively propagated over time by propagating just its mean 𝒙𝒙 and covariance 𝑷𝑷 A multivariate Gaussian distribution 𝒩𝒩 𝑥𝑥, 𝑃𝑃 is completely characterized by its vector mean and covariance matrix: 𝑥𝑥 ≝ 𝐸𝐸[𝑥𝑥] 𝑃𝑃 ≝ 𝐸𝐸 𝑥𝑥 − 𝑥𝑥 𝑥𝑥 − 𝑥𝑥 𝑇𝑇 L27 - Smart Systems In other words, if we can solve for equations to propagate the mean and covariance through time (at each step), we can come up with a solution to the linear estimation problem 26 27.1 Course Review Smart Systems Measurement 𝑥𝑥 Estimated Output 𝑧𝑧̂ 𝑧𝑧 Unrefined Estimated States Prediction: Correction Term 𝑥𝑥 = 𝐴𝐴𝑥𝑥 + 𝐵𝐵𝐵𝐵 Input 𝐴𝐴, 𝐵𝐵 or Gain 𝑧𝑧̂ = 𝐶𝐶 𝑥𝑥 L27 - Smart Systems Update: 𝑢𝑢 System Model 𝐾𝐾 = ? 𝑥𝑥 = 𝑥𝑥 + 𝐾𝐾 𝑧𝑧 − 𝑧𝑧̂ Refined 𝑥𝑥 Estimates 𝐾𝐾 27 27.1 Course Review Smart Systems Prediction stage: 𝑥𝑥 𝑘𝑘+1|𝑘𝑘 = 𝐴𝐴𝑥𝑥 𝑘𝑘|𝑘𝑘 + 𝐵𝐵𝑢𝑢𝑘𝑘 𝑃𝑃𝑘𝑘+1|𝑘𝑘 = 𝐴𝐴𝑃𝑃𝑘𝑘|𝑘𝑘 𝐴𝐴𝑇𝑇 + 𝑄𝑄𝑘𝑘 Update stage: 𝑆𝑆𝑘𝑘+1 = 𝐶𝐶𝑃𝑃𝑘𝑘+1|𝑘𝑘 𝐶𝐶 𝑇𝑇 + 𝑅𝑅𝑘𝑘+1 −1 𝐾𝐾𝑘𝑘+1 = 𝑃𝑃𝑘𝑘+1|𝑘𝑘 𝐶𝐶 𝑇𝑇 𝑆𝑆𝑘𝑘+1 L27 - Smart Systems 𝑥𝑥 𝑘𝑘+1|𝑘𝑘+1 = 𝑥𝑥 𝑘𝑘+1|𝑘𝑘 + 𝐾𝐾𝑘𝑘+1 𝑧𝑧𝑘𝑘+1 − 𝐶𝐶 𝑥𝑥 𝑘𝑘+1|𝑘𝑘 𝑇𝑇 𝑃𝑃𝑘𝑘+1|𝑘𝑘+1 = 𝐼𝐼 − 𝐾𝐾𝑘𝑘+1 𝐶𝐶 𝑃𝑃𝑘𝑘+1|𝑘𝑘 𝐼𝐼 − 𝐾𝐾𝑘𝑘+1 𝐶𝐶 𝑇𝑇 + 𝐾𝐾𝑘𝑘+1 𝑅𝑅𝑘𝑘+1 𝐾𝐾𝑘𝑘+1 Note: Sometimes we can update the covariance as follows, but it is not as numerically stable: 𝑃𝑃𝑘𝑘+1|𝑘𝑘+1 = 𝐼𝐼 − 𝐾𝐾𝑘𝑘+1 𝐶𝐶 𝑃𝑃𝑘𝑘+1|𝑘𝑘 28  Notation Definition Dimensions 𝑛𝑛: # of states 𝐴𝐴 System matrix 𝑛𝑛 × 𝑛𝑛 𝑚𝑚: # of measurements 𝑟𝑟: # of inputs 𝐵𝐵 Input gain matrix 𝑛𝑛 × 𝑟𝑟 27.1 Course Review 𝐶𝐶 Measurement matrix 𝑚𝑚 × 𝑛𝑛 𝑥𝑥 State 𝑛𝑛 × 1 𝑧𝑧 Measurement 𝑚𝑚 × 1 𝑢𝑢 Input 𝑟𝑟 × 1 𝑥𝑥 Estimated state 𝑛𝑛 × 1 𝑧𝑧̂ Estimated measurement 𝑚𝑚 × 1 𝐼𝐼 Identity matrix 𝑛𝑛 × 𝑛𝑛 Nomenclature 𝐾𝐾 Gain matrix (Kalman) 𝑛𝑛 × 𝑚𝑚 𝑃𝑃 State error covariance matrix 𝑛𝑛 × 𝑛𝑛 L27 - Smart Systems 𝑄𝑄 System noise covariance 𝑛𝑛 × 𝑛𝑛 𝑅𝑅 Measurement noise covariance 𝑚𝑚 × 𝑚𝑚 𝑆𝑆 Innovation covariance 𝑚𝑚 × 𝑚𝑚 𝑇𝑇 Sample rate 1×1 𝑘𝑘 Time step 1×1 𝑘𝑘 + 1|𝑘𝑘 Prediction step (subscript) 1×1 𝑘𝑘 + 1|𝑘𝑘 + 1 Update step (subscript) 1×1 29 27.1 Course Review Smart Systems Artificial intelligence (AI) is computer software that mimics human cognitive abilities to perform complex tasks historically done by humans It is an umbrella term covering a variety of interrelated, but distinct, subfields  Machine learning: a subset of AI on which algorithms are trained on datasets to become capable of performing specific tasks L27 - Smart Systems  Deep learning: a subset of ML, in which artificial neural networks that mimic the brain are used to perform more complex reasoning tasks 30 27.1 Course Review Smart Systems Linear regression is a supervised technique used for predicting continuous target variables based on one or more input features Simple linear regression involves predicting a target variable 𝒚𝒚 using a single predictor 𝒙𝒙 with a linear equation: 𝑦𝑦 = 𝛽𝛽0 + 𝛽𝛽1 𝑥𝑥 L27 - Smart Systems 𝛽𝛽0 and 𝛽𝛽1 are known as weights, or regression coefficients 31 27.1 Course Review Smart Systems Simple linear regression considers one input feature, whereas multiple linear regression can consider more than one input feature Cost functions are used to evaluate the accuracy of a model’s predictions The gradient descent algorithm is used to update the model’s parameters or weights based on the cost function L27 - Smart Systems Testing sets are used to evaluate a model’s performance after training 32 27.1 Course Review Smart Systems Deep learning is a subset of machine learning and involves a group of techniques based on neural networks Neural networks have the capacity to learn complex patterns directly from the data These models have three types of layers:  Input layer: where the input data is introduced L27 - Smart Systems  Hidden layer: responsible for learning patterns from data  Output layer: responsible for outputting a prediction 33 27.1 Course Review Smart Systems CNNs are neural networks which are specialized in dealing with image data They make use of several new types of layers:  Convolutional layer  Pooling layer The added layers in CNNs make use L27 - Smart Systems of filters which can capture local patterns in the input, making them highly effective for image analysis 34 27.1 Course Review Smart Systems Recurrent neural networks (RNNs) are another type of deep learning model which are specifically designed for processing sequences of data RNNs have recurrent connections, which allow them to pass information from one step in the sequence to the next  This looping mechanism enables RNNs to capture temporal dependencies in the data L27 - Smart Systems At each time step, RNNs maintain a hidden state  Representation of information seen up to that point in the sequence 35 27.1 Course Review Smart Systems Generative AI models typically rely on having access to a vast amount of data, depending on the application The GPT models that power ChatGPT have been trained on millions if not hundreds of millions of text data  This includes sheet music, poetry, code/programs, books, research, and more! L27 - Smart Systems Similarly, Deepfake models have been trained on vast amounts of data consisting of images of objects, animals, and people There are many pros and cons to GenAI… 36 27.2 Course Review Midterm Reminders Midterm is Thursday, November 21st in HH 305 at 1:30 pm You will need pens, pencils, calculator and your student ID  No formula/equations sheet required It is worth 20% of your final grade There are four main questions, plan your time accordingly! L27 - Smart Systems 37 27.3 Course Review Wrap-up This course provided an introduction to the fundamentals of smart systems. By the end of it, you should will be able to: 1. Mathematically model mechanical and/or electrical systems through first-order equations and state-space representation. 2. Simulate linear and nonlinear dynamic systems in MATLAB or Python. L27 - Smart Systems 3. Explain estimation theory and derive the Kalman filter equations. 4. Program, simulate, and implement the Kalman filter in MATLAB or Python. 5. Explain and tune PID controllers in both a simulated environment (MATLAB or Python) and embedded system. 38 27.3 Course Review Wrap-up …and: 6. Understand how to properly collect and process data from sensors on an experimental setup. 7. Understand how to apply machine learning strategies on data collected from an experimental setup for optimization purposes. 8. Properly format and prepare a resume for a relevant technical job L27 - Smart Systems posting found in industry. 39 27.3 Course Review Wrap-up Breakdown of course grade:  Assignments: 30%  Four assignments in total (individual)  Resume Workshop: 5%  Career services will assist... (individual) L27 - Smart Systems  Midterm: 20% (individual)  Project: 45%  Project ‘day’ is usually Thursday (group)  Due early December (but it will sneak up on you!) 40 27.3 Course Review Wrap-up Thank you for enrolling in Smart Systems It was a second offering at McMaster, and we hope that you have enjoyed it (for the most part!) We are open to suggestions for improving the course, as well as other general feedback on things you liked and did not like  Email reminders will be sent out later this semester about course L27 - Smart Systems evaluations/feedback After graduation (assuming you pass ME 4SS3…), please feel free to reach out and stay in touch! 41  Monday Wednesday Thursday Deliverables (Virtual on A2L or MS Teams) (HH 305) (HH 305) - - L01: Introduction 09/04 L02: Introduction to 09/05 - to Course Project L03: Introduction to 09/09 L04: System 09/11 L05: Project - Laser 09/12 - Smart Systems Modeling I Cutting (JHE A104A) L06: System Modeling 09/16 L07: System Modeling 09/18 L08: Project - Assembly 09/19 - Practice II Check and Testing L09: System Modeling 09/23 L10: System Modeling 09/25 L11: Project - Modeling 09/26 Assignment 1 and Matlab I and Matlab II Examples (09/29) No Class 09/30 L12: Control Theory 10/02 L13: Project - Matlab 10/03 - (Truth and Simulation Reconciliation) L14: Signal 10/07 L15: Controllers 10/09 L16: Project - 10/10 - Conditioning and Matlab Arduino Tutorial What’s next? No Class 10/14 No Class 10/16 No Class 10/17 Assignment 2 (Thanksgiving Monday) (Break) (Break) (10/20) L17: Introduction to 10/21 L18: Kalman Filter I 10/23 L19: Project – TA 10/24 - Estimation Theory Consultations L20: Kalman Filter II 10/28 L21: Kalman Filter 10/30 L22: Project - TA 10/31 Assignment 3 L27 - Smart Systems and Matlab Consultations (11/03) L23: Introduction to 11/04 L24: Resume 11/06 L25: HR Industry 11/07 Assignment 4 Machine Learning Workshop (Online) Panel (In-Person) (11/10) L26: Machine Learning 11/11 L27: Review of Smart 11/13 L28: Project - TA 11/14 Resume Asgmt. Applications Systems Material Consultations (11/17) Virtual Office Hours on 11/18 L29: Midterm Review 11/20 L30: Midterm 11/21 Midterm MS Teams and Help (11/21) Virtual Office Hours on 11/25 L31: Project - TA 11/27 L32: Project - 11/28 Project Demo MS Teams Consultations (11/28) 42 Demonstration Day Virtual Office Hours on 12/02 Virtual Office Hours on 12/04 Project - Report Due 12/05 Project Report MS Teams MS Teams (No Class) (12/05) Additional Resources Smart Systems Slides (PDF) and relevant code will be found on the course page within Avenue to Learn Relevant textbooks and resources are referenced in the syllabus or within the slides Please contact me if you have any difficulties throughout the course L27 - Smart Systems 43