Course detail

Control of Linear Time-Invariant Systems

FSI-VA1 Acad. year: 2027/2028 Winter semester

The course follows on from Theory of Automatic Control (VZR) and extends the classical approach to control with modern design methods based on the state-space description. It focuses on the control design of dynamic systems in continuous and discrete time and its application in engineering practice. It is built on linear time-invariant systems and extends to the control of nonlinear systems and learning-based methods.

The course covers the state-space description, controllability and observability, design of state controllers and observers, linear quadratic optimal control (LQR, LQG), robust control, control of multivariable systems, model predictive control (MPC), control of nonlinear systems using linearisation and Takagi-Sugeno fuzzy models, and learning-based control methods (neural networks, learning MPC, reinforcement learning). The course concludes with control implementation in practice.

The core of the exercises is a team project in which students design, implement and verify control of a practical application. Students first master the methods covered in MATLAB/Simulink and then apply them in the project.

Language of instruction

Czech

Number of ECTS credits

6

Entry knowledge

The student has mastered classical control theory to the extent of the course Theory of Automatic Control (VZR): transfer functions, time and frequency responses, stability and control performance, PID controller tuning, cascade control and discretisation. The student can work with matrices including eigenvalues and use MATLAB/Simulink.

Rules for evaluation and completion of the course

Course-unit credit (max. 30 points): a team project – design, implementation and verification of control of a practical application, with an interim presentation and a final defence; both the team result and the individual contribution of the student are assessed. The credit requires at least 15 points (50 %) from the project.

Examination (max. 70 points): oral; questions on the design and interpretation of state-space, optimal and predictive control and on their connection with the team project. The examination is passed with at least 35 points (50 %).

Final grade: sum of the project and examination points (max. 100) on the ECTS scale per Art. 14 of the BUT Study and Examination Rules: A 90–100, B 80–89, C 70–79, D 60–69, E 50–59, F 0–49.

Attendance: attendance at lectures and exercises is compulsory and checked by the tutor. One absence from an exercise can be made up by attending another group in the same week or by a substitute task; a longer absence is made up by written substitute tasks set by the tutor. Unexcused absence means the credit is not awarded.

Aims

On completing the course, the student is able to:

  1. build a state-space model of an engineering system convert it between the external and internal description and assess its controllability and observability;
  2. design a state controller by pole placement including integral action and disturbance compensation;
  3. design a state observer and a Kalman filter and incorporate them into the control loop;
  4. design an LQR/LQG controller and a model predictive controller (MPC) with constraints, justify the choice of their parameters and assess the potential of learning-based methods (neural networks, learning MPC, reinforcement learning);
  5. assess the robustness of a design against model uncertainty and design control of a multivariable system and of a nonlinear system using linearisation and a Takagi-Sugeno fuzzy model;
  6. design, implement and verify control of a practical application as a team and defend the chosen solution.

Study aids

Lecture slides, exercise assignments and MATLAB/Simulink models, and team project assignments are available in the course on the BUT e-learning (Moodle).

The study programmes with the given course

Programme N-AIŘ-P: Applied Computer Science and Control, Master's, compulsory

Type of course unit

 

Lecture

39 hours, compulsory

Syllabus


  1. From the transfer function to the state-space description: state-space model, conversion between external and internal description, canonical forms.

  2. Solution of state equations, stability, discrete-time state-space model. Controllability and observability.

  3. State feedback and pole placement, integral action, reference tracking, disturbance compensation.

  4. State observer and the Kalman filter, separation principle, controller with an observer.

  5. Linear quadratic optimal control: LQR and LQG, choice of weights.

  6. Model predictive control (MPC): principle, constraints, tuning.

  7. Robustness and robust design: model uncertainty, sensitivity function, stability margins, fundamentals of robust synthesis.

  8. Multivariable systems: coupling, autonomy, decoupling. Transport delay compensation.

  9. Control of nonlinear systems: linearisation at an operating point, gain scheduling, verification on the nonlinear model.

  10. Fuzzy approach to nonlinear control: the Takagi-Sugeno model as a combination of local linearisations, a fuzzy controller composed of local controllers.

  11. Neural networks in control: a neural network as a plant model (including the physics-informed approach), neural network approximation of MPC.

  12. Learning from data in control: learning MPC, fundamentals of reinforcement learning and its relation to optimal control, maintaining constraints and safety.

  13. Control implementation in practice: discrete realisation of the controller and observer, actuator limits, deployment of the controller on embedded hardware, tuning on a real plant.

Computer-assisted exercise

26 hours, compulsory

Syllabus


  1. State-space model of a plant in MATLAB/Simulink, conversion between descriptions, controllability and observability.

  2. Pole placement, state controller with integral action and disturbance compensation.

  3. State observer and the Kalman filter.

  4. Team project: assignment, team roles.

  5. Team project: plant model and its validation.

  6. Team project: control design (state controller, LQR).

  7. Team project: model predictive control (MPC) as a design variant.

  8. Team project: comparison of control variants and robustness assessment.

  9. Team project: interim presentation.

  10. Team project: implementation.

  11. Team project: neural network controller or learning MPC as an optional variant.

  12. Team project: tuning and verification.

  13. Project defence, credit.