Awesome Control Theory

A map-aligned guide to control theory topics, inspired by Brian Douglas's Map of Control Theory © August 2020, Engineering Media.

Control theory is the study of dynamical systems that sense, decide, and act so their behavior matches a desired objective. This repository turns the visual map into a structured reference for learning, review, and project planning.

How to Use This Map

Use Modeling and Simulation to describe the plant.
Use System Analysis to understand stability, performance, and limitations.
Use Control Methods to choose a controller family.
Use Planning to generate references, paths, and constraints.
Use State Estimation when the full state is not directly measurable.
Use First Principles and Classical Tools to connect physical models, diagrams, and approximations to controller design.

Core Feedback Concepts

Feedback and Feedforward

Feedback control closes the loop by measuring behavior and correcting error.
Feedforward control acts from a model or known command before error appears.
Positive feedback reinforces deviations and is useful in oscillators, switches, regenerative amplification, and some biological circuits, but it usually needs nonlinear saturation or logic to avoid runaway behavior.
Feedback with logic combines continuous loops with selectors, modes, interlocks, startup/shutdown sequences, and supervisory decisions.
Control-system architectures organize feedback, feedforward, estimation, reference generation, optimization, and logic into nested or layered designs.
Full-state feedback uses the entire state vector, often in the form u = -Kx, when states are measured or estimated.
Output feedback uses measured outputs directly or through an observer.
Separation principle designs state feedback and an observer separately for many linear output-feedback problems.
Compensator design combines controller and observer dynamics when the full state is not directly measured.

Domains and Representations

Continuous time models use differential equations such as dx/dt = f(x, u).
Discrete time models use sampled updates such as x[k+1] = f(x[k], u[k]).
Time-domain analysis studies transients, step response, impulse response, rise time, settling time, overshoot, and steady-state error.
Frequency-domain analysis studies gain, phase, bandwidth, resonances, and robustness using transfer functions and frequency response plots.
Laplace transforms connect differential-equation models to transfer functions and classical design tools.
C2D and D2C conversion translate models between continuous and discrete representations for sampled-data control.

Mathematical and Computational Foundations

Linear algebra underpins state-space models, controllability, observability, pole placement, and model reduction.
Differential equations describe continuous-time plant dynamics and closed-loop behavior.
Convex optimization supports many MPC, estimation, and robust-control formulations.
Dynamic programming connects optimal control, HJB equations, value functions, and reinforcement learning.
Riccati equations appear in LQR, LQG, Kalman filtering, and finite-horizon optimal control.
Quadratic programming is the standard online optimization form for many linear MPC and control-barrier-function controllers.
Probability and stochastic processes support Kalman filtering, stochastic control, random disturbances, covariance propagation, and spectral-density descriptions of noise.
Numerical optimization supports direct optimal control, trajectory optimization, constrained MPC, parameter estimation, and design tuning.

Control Methods

Linear

PID control combines proportional, integral, and derivative action for simple, effective feedback control.
PID tuning chooses proportional, integral, and derivative gains using model-based rules, relay experiments, frequency response, or empirical iteration.
Integral action and anti-windup improve steady-state tracking while limiting integrator problems when actuators saturate.
Lead-lag compensation shapes transient response and steady-state accuracy with phase-lead and phase-lag networks.
Full-state feedback places closed-loop poles through a gain matrix K.
Pole placement assigns closed-loop eigenvalues directly when the model is controllable.
LQR, or Linear Quadratic Regulator, minimizes a quadratic state and input cost to produce an optimal state-feedback controller.
LQG, or Linear Quadratic Gaussian control, combines LQR with Kalman filtering for noisy output-feedback problems.
H-infinity control is often synthesized for linear plants while targeting worst-case disturbance attenuation and robustness.
Loop shaping designs open-loop gain and phase to meet bandwidth, tracking, noise, and robustness targets.
Loop-transfer-function design studies L = PC as the object that ties stability margins, sensitivity, complementary sensitivity, and bandwidth together.

Nonlinear

Gain scheduling blends or switches controllers across operating points.
Backstepping recursively designs controllers for strict-feedback nonlinear systems.
Feedback linearization cancels nonlinearities through feedback to obtain a simpler closed-loop form.
Dynamic inversion uses a model inverse to command nonlinear systems.
Sliding mode control drives trajectories onto a designed switching surface and provides robustness to matched uncertainty.
Bang-bang control switches between extreme control values, often appearing in minimum-time problems.
Describing functions, equivalent gains, and the circle criterion provide approximate and absolute-stability tools for nonlinear feedback systems.
Inverse nonlinearities and time-optimal servo structures compensate known nonlinearities or exploit actuator limits for fast motion.
Perturbation, averaging, and singular-perturbation methods analyze systems with weak nonlinearities, periodic effects, or separated time scales.
Input-output stability and passivity provide nonlinear feedback-analysis tools that complement Lyapunov methods.

Multi-Agent

Graph-theoretic control models agents and communication links as nodes and edges.
Consensus control drives distributed agents toward agreement using local communication.
Formation control regulates relative positions, distances, or bearings among agents.
Leader-follower control coordinates agents around one or more reference agents.
Swarm control produces group behavior from decentralized local rules.

Optimal

Pontryagin's Maximum Principle converts optimal-control problems into necessary conditions involving a Hamiltonian and costates.
Hamilton-Jacobi-Bellman equation expresses the dynamic-programming condition for optimal feedback control.
LQR is the canonical linear-quadratic optimal controller.
Differential dynamic programming and iterative LQR approximate nonlinear optimal-control problems through local quadratic models.
Direct collocation and shooting methods transcribe continuous trajectory optimization into finite-dimensional nonlinear programs.
Bang-bang solutions can arise when the optimal input saturates at limits.
Calculus of variations gives Euler-Lagrange-style necessary conditions for optimal trajectories.
Dynamic-programming algorithms solve shortest-path, finite-horizon, infinite-horizon, perfect-information, and imperfect-information decision problems.
Stochastic optimal control handles dynamics, measurements, or disturbances modeled probabilistically.
Dual control accounts for the fact that inputs can both control the plant and excite it to learn uncertain parameters.

Predictive

Model Predictive Control, or MPC, repeatedly solves a finite-horizon optimization problem using a model of the plant.
Linear MPC uses linear models and often convex quadratic programs.
Nonlinear MPC handles nonlinear dynamics or constraints, usually through nonlinear programming.
Robust MPC accounts for model uncertainty and disturbances.
Tube MPC keeps the uncertain state inside a planned invariant tube around a nominal trajectory.
Stochastic MPC treats uncertainty probabilistically through chance constraints or expected costs.
Explicit MPC precomputes piecewise-affine control laws for fast online evaluation.
Constraint handling is central to MPC because input, state, and safety limits can be represented directly in the optimization.
Generalized Predictive Control, or GPC, is an influential predictive control strategy based on input-output models and receding-horizon design.
Fast MPC methods exploit structure, warm starts, explicit solutions, or tailored solvers to meet real-time deadlines.
Hybrid MPC handles systems with both continuous dynamics and discrete modes or logic.
Industrial and commercial MPC schemes emphasize model maintenance, constraint management, estimator integration, and reliable online optimization.

Digital and Sampled-Data

Digital control designs controllers that run on sampled measurements and update actuators at discrete instants.
Z-transform methods analyze discrete-time transfer functions and sampled systems.
Difference equations describe recursive discrete-time dynamics directly and connect digital filters to transfer-function realizations.
Pulse transfer functions represent sampled-data input-output behavior at the sampling instants.
Modified z-transforms help analyze delayed and intersample behavior in sampled-data systems.
Bilinear and w-transform methods map discrete-time design questions into continuous-like algebraic forms.
Sampled-data models describe the combined behavior of continuous plants, samplers, zero-order holds, and digital controllers.
Ideal sampling and reconstruction explain spectra, aliasing, and how sampled signals are turned back into continuous actuator commands.
Zero-order hold discretization is the common model for digital-to-analog actuation between samples.
First-order, fractional-order, and triangle holds model alternate data extrapolation assumptions between samples.
Discrete equivalents convert continuous plants or controllers into discrete models using numerical integration, zero-pole matching, hold equivalents, or cost-equivalent emulation.
Emulation design discretizes a continuous-time controller, then checks the sampled closed-loop response.
Direct digital design designs the controller in the z-plane using root locus, frequency response, pole placement, deadbeat control, or Ragazzini methods.
Deadbeat and modified deadbeat control seek finite-sample settling while managing actuator effort and robustness.
Sample-rate selection balances bandwidth, smoothness, aliasing, computation, measurement noise, and sensitivity to plant uncertainty.
Multirate and nonsynchronous sampling handle systems where sensors, actuators, or controllers update at different rates or phases.
Intersample ripple captures output behavior between sampling instants, which can be missed by purely discrete-time analysis.
Aliasing, quantization, and computation delay are practical effects that can change closed-loop behavior.
Round-off, word-length effects, limit cycles, and dither matter when controllers are implemented with finite-precision arithmetic.
A/D and D/A conversion connect sensors, actuators, computers, and continuous plants in practical digital control loops.
Digital controller realization covers direct, parallel, cascade, factorized, and state-space implementations.
Embedded and distributed implementation covers processor interfaces, communication links, reliability, scheduling, and integration with plant hardware.
Hybrid control combines continuous dynamics with discrete control logic, switching, events, or mode-dependent controllers.
Controller implementation covers realizable filters, derivative roll-off, bumpless transfer, saturation handling, PLC or computer deployment, and numerical details that change real closed-loop behavior.

MIMO and Multivariable

MIMO control handles plants with multiple inputs and multiple outputs, where loops can interact strongly.
Large-scale system control uses decomposition, hierarchical control, multilayer coordination, and optimization to manage systems too large for a single centralized loop.
Singular-value analysis studies multivariable gain, directionality, and robustness across frequency.
Relative gain array, or RGA, helps evaluate input-output pairings and control-structure choices.
Decoupling control reduces cross-channel interactions when the model supports it.
Control structure design chooses manipulated variables, controlled variables, measurements, and loop pairings.
Linear Matrix Inequalities, or LMIs, express many robust, optimal, and constrained control conditions as convex feasibility problems.

Intelligent

Fuzzy control uses membership functions and rule bases to encode heuristic control behavior.
Reinforcement learning learns policies from reward through exploration and exploitation.
Adaptive dynamic programming approximates dynamic programming for systems where exact HJB solutions are intractable.
Neural network control uses learned function approximators for policies, dynamics, value functions, or adaptive compensation.
Genetic algorithms tune controller parameters or search design spaces with evolutionary optimization.

Adaptive

Model Reference Adaptive Control, or MRAC, adapts controller parameters so the plant follows a desired reference model.
Direct adaptive control adjusts controller parameters directly from tracking error.
Indirect adaptive control estimates plant parameters first, then updates the controller from the estimated model.
Self-tuning regulators repeatedly identify a model and redesign controller gains online.
Extremum seeking optimizes an unknown objective online by perturbing inputs and following performance gradients.
Iterative Learning Control, or ILC, improves repeated-task tracking from trial to trial.
Auto-tuning and relay feedback identify useful process information online to tune controllers such as PID loops.
Real-time parameter estimation updates model parameters from streaming data during operation.
Robust adaptive control adds safeguards so adaptation remains stable under unmodeled dynamics, noise, and disturbances.

Robust

Active Disturbance Rejection Control, or ADRC, estimates and compensates total disturbances in real time.
H-infinity control minimizes worst-case disturbance amplification.
Mu-synthesis handles structured uncertainty in robust-control design.
Small-gain reasoning bounds feedback interconnections by limiting loop gain under uncertainty.
Linear fractional transformations, or LFTs, organize uncertain plants for robust-analysis and synthesis workflows.
Loop shaping and robust stability margins connect classical frequency design to robust-control goals.
Robust performance analysis verifies both stability and performance under uncertainty.
Structured singular value analysis quantifies robustness for structured uncertainty models.

Planning

Reference Inputs

Step input tests tracking, settling behavior, overshoot, and steady-state error.
Impulse input reveals natural dynamics and impulse response.
Sine input probes frequency response and periodic tracking.

Optimal Planning

Cost functions encode objectives such as time, distance, energy, comfort, risk, or tracking error.
Trajectory optimization searches over state and input histories that satisfy dynamics and minimize cost.
Trajectory generation creates dynamically feasible reference states, velocities, and accelerations for downstream tracking controllers.
Minimum-snap and polynomial trajectories are common in aerial robotics and other systems with smoothness requirements.

Constraints

Input constraints bound actuator effort, rate, torque, force, voltage, acceleration commands, or steering.
State constraints bound position, velocity, acceleration, temperature, pressure, charge, safety envelopes, or operating regions.
Environmental constraints encode obstacles, keep-out zones, and workspace limits.
Collision-avoidance constraints preserve separation from obstacles, humans, vehicles, or other agents.
Terminal constraints enforce desired final states or invariant terminal sets in predictive planning.

Motion Planning

Holonomic systems can move freely in all configuration directions.
Nonholonomic systems have velocity constraints, such as car-like robots.
Redundant systems have more degrees of freedom than needed for the task.

Path Planning Algorithms

RRT, or Rapidly-exploring Random Tree, samples configuration space to find feasible paths in high-dimensional problems.
RRT-star extends RRT with asymptotic optimality.
PRM, or Probabilistic Roadmap, builds a reusable graph of sampled configurations.
Dijkstra's algorithm finds shortest paths on weighted graphs without a heuristic.
A-star (A)* searches graphs with a heuristic to find low-cost paths.

State Estimation

Filtering Methods

Kalman filter estimates linear Gaussian systems optimally in the least-squares sense.
Extended Kalman filter linearizes nonlinear dynamics and measurement models locally.
Sigma-point filters, including the unscented Kalman filter, propagate selected sample points through nonlinear models.
Particle filters approximate arbitrary state distributions with weighted samples.
Information filters represent uncertainty with information matrices and are useful in some distributed or sparse estimation problems.
Square-root and U-D filters propagate covariance factors to improve numerical conditioning.
H-infinity filters estimate states under worst-case disturbance models instead of relying only on stochastic noise assumptions.
Kalman-Bucy filters are continuous-time Kalman filters for linear systems driven by continuous-time stochastic models.
Constrained filters enforce known bounds or equality constraints on state estimates.
Smoothers, such as fixed-lag or Rauch-Tung-Striebel smoothers, estimate past states using measurements that arrived later.
Wiener filtering estimates signals from noisy measurements using second-order statistical descriptions.
Recursive least squares estimates fixed or slowly varying parameters from streaming data.
Covariance tuning and consistency checks keep filter uncertainty aligned with observed residuals.

Observers

State observers reconstruct unmeasured states from models and output measurements.
Luenberger observers use linear correction dynamics.
Disturbance observers estimate unmeasured disturbances for compensation.
Kalman observers combine model prediction and measurement correction with explicit noise statistics.
Prediction and current observers distinguish whether estimates are formed before or after incorporating the newest sampled measurement.
Unknown-input observers estimate states when some disturbances or inputs are not measured.
High-gain and sliding-mode observers are common nonlinear observer families.
Reduced-order observers estimate only the unmeasured portion of the state when some states are directly measured.

Moving Horizon Estimation

Moving Horizon Estimation, or MHE, estimates states and parameters by solving a constrained optimization problem over a recent time window.

Calibration

Bias calibration estimates offsets such as yp = y + b.
Gain and alignment calibration estimates sensor scale factors and mounting geometry.
Parameter calibration fits model coefficients from measured data.

Mapping

Mapping estimates environmental structure for navigation, localization, and planning.
SLAM-style workflows combine mapping with state estimation when position and environment are both uncertain.

Tracking

Target tracking estimates moving object states from noisy measurements.
Trajectory tracking estimates deviation from a desired path or reference.
Multi-sensor tracking fuses detections from multiple measurement sources.
Data association matches measurements to tracks in multi-target tracking problems.
Multiple-model estimation runs several candidate models or filters in parallel to handle mode changes or uncertain dynamics.

Sensor Fusion

IMU, GPS, and camera fusion combines inertial, satellite, and visual measurements for pose and navigation.
Redundant measurement fusion improves reliability and fault tolerance.
Outlier rejection and fault detection prevent bad measurements from corrupting state estimates.
Bayesian filtering provides a common probabilistic foundation for Kalman, sigma-point, and particle filters.

Modeling and Simulation

State-Space Models

Linear state space
```
x_dot = A x + B u
y     = C x + D u
```
Nonlinear state space
```
dx/dt = f(x, u)
y     = g(x, u)
```
Hybrid systems combine continuous dynamics with discrete modes, events, or logic.
Discrete-time state space represents sampled dynamics with updates such as x[k+1] = A x[k] + B u[k].
Matrix-exponential discretization computes exact linear sampled models under zero-order-hold assumptions when the continuous model is known.
Discrete models with delays represent sensor, actuator, communication, and computation delays as augmented states or delayed inputs and outputs.
Reference-input and state-command structures connect state feedback and observers to tracking commands rather than only regulator problems.
State augmentation for integral action adds integrator states or disturbance estimates to remove steady-state errors.
Time-delay systems model transport delays, communication delays, and computation delays that can destabilize feedback loops.
Saturation and rate-limit models capture actuator limits that strongly affect closed-loop performance.
Stochastic state models include process noise, measurement noise, and random disturbances in continuous or discrete time.
Stochastic differential equations model continuous-time dynamics driven by random processes, often using white-noise or Wiener-process idealizations.

System Representations

Transfer functions represent input-output dynamics in the Laplace domain.
Discrete transfer functions represent input-output dynamics in the z-domain.
Block diagrams show interconnections among plants, controllers, summing junctions, sensors, and feedback paths.
Signal-flow views clarify feedback, feedforward, disturbances, and noise.
Signal-flow graphs use node-edge relationships and Mason's gain formula to derive input-output transfer functions.
Uncertainty models describe parametric uncertainty, unmodeled dynamics, disturbances, and sensor noise.
Linear fractional transformations separate nominal dynamics from uncertainty blocks for robust-control analysis.

Model Development

First-principles modeling derives dynamics from physics, such as Newton's laws, energy balances, circuits, fluids, or thermodynamics.
System identification estimates models from input-output data.
Nonparametric identification estimates responses or spectra without first committing to a low-order parametric model.
Parametric identification fits model structures such as transfer functions, state-space models, or input-output regressions.
Black-box identification fits model dynamics from data when the internal physical structure is unknown or intentionally abstracted.
Least-squares, recursive least-squares, stochastic least-squares, maximum likelihood, and subspace identification are common estimation methods for discrete-time models.
Identification experiment design chooses excitation signals, sampling frequency, scaling, and validation data so estimated models are useful for control.
Linearization approximates nonlinear dynamics near an equilibrium or trajectory.
Canonical forms and similarity transformations reorganize state-space models without changing input-output behavior.
Minimum realization removes uncontrollable or unobservable states while preserving input-output behavior.
Model reduction lowers model order for analysis, control synthesis, and real-time simulation.
Balanced truncation reduces stable linear models while approximately preserving input-output behavior.

Simulation

Numerical integration solves ordinary differential equations and differential-algebraic equations.
Event handling captures impacts, switches, guards, saturations, and mode changes in hybrid simulations.
Co-simulation connects tools or subsystem models that use different solvers, time steps, or modeling languages.
Simulation software includes tools such as Simulink, Modelica, and domain-specific simulators for controls, physical modeling, and verification.
Closed-loop simulation validates controller behavior before deployment.
Discrete-time simulation checks difference-equation, digital-filter, and sampled-data controller behavior before hardware implementation.
Monte Carlo simulation tests performance across uncertainty, noise, and randomized initial conditions.
Hardware-in-the-loop simulation exercises controller implementation against real-time plant models.

System Analysis

Stability

Lyapunov stability proves behavior using energy-like functions.
Local stability describes behavior near an equilibrium.
Global stability describes behavior over a broad state space.
Input-to-state stability connects bounded inputs and disturbances to bounded state behavior.
Phase-plane analysis visualizes trajectories, equilibria, limit cycles, and nonlinear behavior in two-state systems.

Margins and Robustness

Gain margin measures allowable gain change before instability.
Phase margin measures allowable phase lag before instability.
Robust stability studies whether stability survives uncertainty.
Robust performance studies whether performance targets survive uncertainty.

Frequency-Domain Analysis

Bode plots show magnitude and phase versus frequency.
Nyquist plots determine closed-loop stability from open-loop encirclements.
Nichols charts combine gain and phase for frequency-domain design.
Sensitivity functions describe disturbance rejection, noise amplification, and tracking limitations.
Nyquist criterion converts encirclements of the critical point by the loop transfer function into a closed-loop stability test.
Bode's relations and waterbed effects explain why reducing sensitivity in one frequency band often increases it elsewhere, especially for nonminimum- phase plants, delays, or unstable poles.
Spectral-density analysis describes how stochastic disturbances and noise are distributed across frequency.

Root Locus and Pole-Zero Analysis

Root locus tracks closed-loop pole movement as gain changes.
Pole-zero plots reveal modes, damping, zeros, cancellations, and nonminimum-phase behavior.
Nonminimum-phase zeros limit tracking speed and transient performance.

Algebraic Stability Tests

Routh-Hurwitz tests determine continuous-time polynomial stability without explicitly computing roots.
Jury tests provide analogous algebraic stability checks for discrete-time characteristic polynomials.
Bilinear-transform stability tests map discrete-time characteristic equations into a form where continuous-time tests can be applied.

System Properties

Controllability tests whether inputs can move the state through the reachable state space.
Observability tests whether outputs contain enough information to recover the state.
Passivity uses energy exchange to reason about stability and interconnections.
Sensitivity describes how references, disturbances, sensor noise, and modeling errors propagate through feedback loops.
Performance includes tracking error, disturbance rejection, noise rejection, bandwidth, overshoot, and control effort.
Steady-state accuracy studies final values, error constants, integral action, and reference/disturbance tracking for continuous and digital loops.
z-plane geometry maps damping, natural frequency, settling behavior, and stability boundaries from the s-plane into sampled systems.
Fundamental limitations identify performance and robustness barriers imposed by right-half-plane poles and zeros, delay, saturation, noise, and actuator bandwidth.

Safety-Critical Analysis

Control Lyapunov functions, or CLFs, encode stabilizing objectives as inequalities.
Control barrier functions, or CBFs, encode forward-invariant safe sets and are often enforced through quadratic programs.
Reachability analysis estimates states that can be reached under dynamics, controls, and disturbances.
Invariant sets define regions where trajectories remain once they enter.
Formal verification checks whether closed-loop behavior satisfies stated safety or temporal-logic requirements.

First Principles and Classical Tools

Mechanical first-principles models include masses, springs, dampers, pendulums, vehicles, and robotics.
Electrical first-principles models include circuits, motors, power converters, and sensors.
Process-control models include tanks, reactors, heat exchangers, distillation columns, and transport delays.
Computing and network models describe queues, admission control, web servers, congestion control, and resource-management loops.
Biological and pharmacokinetic models include gene regulation, physiological feedback, drug-compartment models, and population dynamics.
Robotics and vehicle models include kinematics, rigid-body dynamics, tire/ground interaction, and actuator dynamics.
Aerospace and pointing models include satellite attitude, antenna azimuth, aircraft landing, and servomotor dynamics.
Thrust vector control uses gimbaled or vectored thrust and inner/outer attitude-position loops, as in rockets and VTOL aircraft.
Operational-amplifier and precision-instrument models connect feedback to high-gain electronics, analog controller realization, and atomic-force- microscope nanopositioning.
Power-system models include generator, grid-interconnection, and topology identification examples.
Precision motion and storage models include flexible structures, disk drive servos, voice-coil actuators, runout, and amplifier saturation.
Linearization connects nonlinear first-principles models to linear design workflows.
Transfer functions connect physical equations to classical feedback design.
Block diagrams organize plant, controller, actuator, sensor, reference, disturbance, and noise pathways.
Safety constraints define forbidden states, operating envelopes, and acceptable risk.

Learning Roadmaps

Beginner Classical Path

First-principles modeling
Transfer functions and block diagrams
Step, impulse, and sine responses
PID control, PID tuning, and implementation details
Root locus, Bode plots, Nyquist plots, and stability margins
Lead-lag compensation, loop transfer functions, and loop shaping

State-Space Path

Linear state-space models
Controllability and observability
Full-state feedback and pole placement
LQR, Riccati equations, and LQG
Observers, Kalman filtering, and separation principle
Nested architectures, MPC, constrained control, and safety filters

Robotics and Autonomy Path

Nonlinear state-space models
Robotics, vehicle, and thrust-vector-control models
Motion planning with holonomic and nonholonomic constraints
RRT, A-star (A*), PRM, and trajectory optimization
Sensor fusion with IMU, GPS, and camera measurements
Mapping, tracking, moving horizon estimation, and safety constraints

Digital Control Path

Sampling, reconstruction, aliasing, and hold devices
Difference equations, z-transforms, pulse transfer functions, and z-plane pole-zero geometry
Discrete equivalents: numerical integration, zero-pole matching, and hold equivalents
Digital controller design: emulation, direct z-plane design, deadbeat control, discrete PID, and state-space pole assignment
Sample-rate selection, multirate effects, delays, quantization, round-off, limit cycles, and dither
Digital implementation: A/D and D/A conversion, controller realization, controller implementation, embedded interfaces, reliability, and hardware-in- the-loop validation

Advanced Control Path

Lyapunov stability
Nonlinear control: feedback linearization, backstepping, sliding mode
Optimal control: PMP, HJB, trajectory optimization, DDP, and iLQR
Robust control: H-infinity, mu-synthesis, ADRC, small-gain, and LFTs
Fundamental limitations, Bode relations, and robust performance tradeoffs
Adaptive, multi-agent, safety-critical, intelligent, and learning-based control

Academic and Open References

These links combine university notes, open textbooks, classic papers, well-maintained open software, and canonical textbooks.

Foundational Texts and Course Notes

Feedback Systems: An Introduction for Scientists and Engineers by Karl J. Astrom and Richard M. Murray: open textbook covering modeling, feedback, stability, frequency response, state space, and robustness.
MIT OCW 6.241J Dynamic Systems and Control: graduate course notes on dynamic systems, feedback interconnections, stability, performance, and robust control.
Caltech CDS 101/110: course material aligned with the Astrom-Murray feedback systems text.
Stanford EE363 Linear Dynamical Systems: lecture slides, support notes, homework, and MATLAB files for state-space models, controllability, observability, least squares, and estimation.
Stanford EE364B Convex Optimization II: optimization notes with material on robust optimization, stochastic MPC, model predictive control, and numerical methods.
MIT Underactuated Robotics: open notes on nonlinear dynamics, LQR, trajectory optimization, planning, Lyapunov analysis, learning, and robotics applications.
Model Predictive Control: Theory, Computation, and Design by Rawlings, Mayne, and Diehl: open MPC textbook with theory, algorithms, computation, and examples.

Textbooks and Learning Resources

Feedback Systems: An Introduction for Scientists and Engineers by Karl J. Astrom and Richard M. Murray: broad, open introduction to feedback principles, modeling, linear systems, state and output feedback, frequency design, PID, robust performance, and system architecture.
Feedback Control of Dynamic Systems by Gene F. Franklin, J. David Powell, and Abbas Emami-Naeini: classical and state-space design text covering dynamic models, response, root locus, frequency response, state-space design, digital control, nonlinear systems, and case studies.
Digital Control of Dynamic Systems, 3rd ed. by Gene F. Franklin, J. David Powell, and Michael L. Workman: digital-control reference covering sampled-data modeling, z-transform analysis, discrete equivalents, transform and state-space design, multivariable optimal control, quantization, sample-rate selection, identification, nonlinear effects, and a disk-drive servo case study.
Applied Digital Control: Theory, Design and Implementation by J. R. Leigh: applied digital-control text covering sampling, z-transform methods, root locus and frequency-response design, digital algorithms, sensors and converters, implementation case histories, state-variable methods, large-scale systems, distributed computer control, adaptive control, and robust control.
Digital Control System Analysis & Design, Global Edition by Charles L. Phillips, H. Troy Nagle, and Aranya Chakrabortty: detailed digital-control text covering discrete-time systems, sampling and reconstruction, open- and closed-loop sampled systems, stability analysis, digital controller design, pole assignment, observers, system identification, linear-quadratic control, Kalman filtering, and application case studies.
Control Systems Engineering by Norman S. Nise: undergraduate controls reference covering modeling, time response, subsystem reduction, stability, steady-state error, root locus, frequency response, state-space design, and digital control.
Modern Control Engineering by Katsuhiko Ogata: standard text on control-system modeling, mechanical, electrical, fluid, and thermal systems, transient and steady-state response, root locus, frequency response, PID, and state-space analysis/design.
Modern Control Systems by Richard C. Dorf and Robert H. Bishop: broad modern-control textbook on mathematical models, state variables, feedback characteristics, performance, stability, root locus, frequency-domain methods, robust control, and digital control.
Control System Design by Graham C. Goodwin, Stefan F. Graebe, and Mario E. Salgado: design-oriented text covering feedback principles, SISO and MIMO control, PID, sampled-data control, hybrid control, optimization-based control, state-space methods, nonlinear control, MPC, and decoupling.
Multivariable Feedback Control: Analysis and Design by Sigurd Skogestad and Ian Postlethwaite: advanced reference for SISO and MIMO limitations, uncertainty, robust stability and performance, controller design, control-structure design, model reduction, LMIs, and case studies.
Nonlinear Systems by Hassan K. Khalil: nonlinear-systems reference covering phase-plane behavior, fundamental properties, Lyapunov stability, input-output stability, passivity, perturbation methods, singular perturbations, and feedback linearization.
Model Predictive Control by Eduardo F. Camacho and Carlos Bordons: MPC reference covering generalized, commercial, multivariable, constrained, robust, nonlinear, hybrid, and fast model predictive control methods.
Dynamic Programming and Optimal Control by Dimitri P. Bertsekas: two-volume reference on dynamic programming, deterministic and stochastic decision problems, shortest paths, imperfect state information, infinite-horizon problems, approximate dynamic programming, and continuous-time optimal control.
Optimal Control Theory: An Introduction by Donald E. Kirk: compact introduction to performance measures, dynamic programming, calculus of variations, Pontryagin's minimum principle, and numerical optimal-trajectory methods.
Control System Design: An Introduction to State-Space Methods by Bernard Friedland: state-space design reference covering feedback, dynamic models, frequency-domain analysis, controllability, observability, pole placement, observers, separation principle, LQR, random processes, and Kalman filtering.
Optimal Control and Estimation by Robert F. Stengel: integrated treatment of optimal trajectories, linear-quadratic control, optimal state estimation, Kalman filtering, stochastic optimal control, dual control, and multivariable design.
Optimal State Estimation: Kalman, H-infinity, and Nonlinear Approaches by Dan Simon: estimation-focused reference covering least squares, Kalman filters, information and square-root forms, smoothing, H-infinity filtering, extended and unscented Kalman filters, and particle filters.
Adaptive Control by Karl J. Astrom and Bjorn Wittenmark: adaptive-control reference covering real-time parameter estimation, self-tuning regulators, MRAS, adaptive-system properties, stochastic adaptive control, auto-tuning, gain scheduling, and implementation.
Introduction to Stochastic Control Theory by Karl J. Astrom: stochastic-control text covering stochastic processes, stochastic state models, spectral descriptions, stochastic differential equations, parametric optimization, and optimal stochastic control.
Schaum's Outline of Feedback and Control Systems by Joseph J. DiStefano III, Allen R. Stubberud, and Ivan J. Williams: problem-oriented review of terminology, differential and difference equations, Laplace and Z-transforms, stability, transfer functions, block diagrams, signal-flow graphs, Nyquist, root locus, Bode, Nichols, nonlinear control, and advanced topics.
Steve Brunton's Control Bootcamp: YouTube playlist introducing control-system modeling, analysis, and design with practical examples.
Brian Douglas Control Systems Lectures: YouTube channel with approachable lectures on classical control, frequency-domain methods, state-space control, and control intuition.
MathWorks/MATLAB YouTube Channel: videos on MATLAB, Simulink, Control System Toolbox workflows, modeling, simulation, and control design examples.
Prof Giordano Scarciotti YouTube Channel: lecture videos on control theory, dynamical systems, and related engineering mathematics.
Robotic Systems Control YouTube Channel: videos on robotics-oriented control, system modeling, estimation, and implementation topics.

Classic Papers and Surveys

Kalman, "A New Approach to Linear Filtering and Prediction Problems": original Kalman filtering paper.
Mayne, Rawlings, Rao, and Scokaert, "Constrained Model Predictive Control: Stability and Optimality": core MPC stability and optimality reference.
LaValle, "Rapidly-Exploring Random Trees: A New Tool for Path Planning": original RRT technical report.
Hart, Nilsson, and Raphael, "A Formal Basis for the Heuristic Determination of Minimum Cost Paths": original A* paper.
Ames et al., "Control Barrier Functions: Theory and Applications": survey-style reference on safety-critical control with CBFs.
Xu, Tabuada, Grizzle, and Ames, "Robustness of Control Barrier Functions for Safety Critical Control": robustness and CLF-CBF quadratic-program formulations.

Open Software and Benchmarks

python-control: Python library for classical and state-space control analysis and design.
Drake: model-based robotics toolbox with simulation, optimization, planning, and control tools.
CasADi: symbolic framework for nonlinear optimization and optimal control.
do-mpc: Python toolbox for nonlinear MPC and moving horizon estimation.
JuliaControl: Julia ecosystem for control systems modeling, analysis, and synthesis.

Coverage Checklist

This checklist mirrors the labels in the map so gaps are easy to spot.

Control methods: linear, nonlinear, multi-agent, optimal, predictive, digital, sampled-data, MIMO, intelligent, adaptive, robust.
Mathematical foundations: linear algebra, differential equations, convex optimization, numerical optimization, probability, stochastic processes, dynamic programming, Riccati equations, quadratic programming.
Linear methods: PID, integral action, anti-windup, lead-lag, pole placement, full-state feedback, output feedback, separation principle, compensator design, LQR, LQG, H-infinity control, loop transfer functions, PID tuning, loop shaping.
Nonlinear methods: gain scheduling, backstepping, feedback linearization, dynamic inversion, sliding mode, bang-bang, describing functions, equivalent gains, circle criterion, inverse nonlinearities, time-optimal servos, perturbation methods, averaging, singular perturbations, input-output stability, passivity.
Multi-agent methods: graph-theoretic control, consensus, formation control, leader-follower control, swarm control.
Optimal methods: Pontryagin's Maximum Principle, HJB equation, LQR, DDP, iLQR, calculus of variations, direct collocation, shooting methods, dynamic-programming algorithms, stochastic optimal control, dual control, optimal planning.
Predictive methods: MPC, linear MPC, nonlinear MPC, robust MPC, tube MPC, stochastic MPC, explicit MPC, generalized predictive control, fast MPC, hybrid MPC, industrial MPC.
Digital and sampled-data methods: digital control, Z-transform, difference equations, pulse transfer functions, modified z-transform, w-transform, sampled-data models, ideal sampling, reconstruction, zero-order hold discretization, first-order and fractional-order holds, discrete equivalents, emulation design, direct z-plane design, deadbeat control, sample-rate selection, multirate sampling, nonsynchronous sampling, intersample ripple, aliasing, quantization, round-off, word-length effects, limit cycles, dither, computation delay, A/D and D/A conversion, controller realization, controller implementation, embedded and distributed implementation, hybrid control.
MIMO and multivariable methods: MIMO control, singular-value analysis, relative gain array, decoupling control, large-scale system decomposition, hierarchical control, multilayer control, control structure design, LMIs.
Intelligent methods: fuzzy control, reinforcement learning, genetic algorithms, adaptive dynamic programming, neural network control.
Adaptive methods: MRAC, direct adaptive control, indirect adaptive control, self-tuning regulators, extremum seeking, iterative learning control, auto-tuning, relay feedback, real-time parameter estimation, robust adaptive control.
Robust methods: ADRC, H-infinity control, mu-synthesis, small-gain reasoning, LFT uncertainty models, structured singular value analysis, robust margins, robust performance.
Planning: step, impulse, sine, constraints, velocity limits, acceleration limits, force limits, holonomic, nonholonomic, redundant systems, trajectory generation, RRT, RRT-star, PRM, Dijkstra, A-star (A*).
State estimation: filtering, observer design, Kalman filter, information filters, square-root filters, U-D filters, H-infinity filters, Kalman-Bucy filters, constrained filters, sigma-point filters, particle filters, smoothers, Wiener filtering, recursive least squares, covariance tuning, prediction observers, current observers, reduced-order observers, unknown-input observers, moving horizon estimation, calibration, mapping, tracking, multiple-model estimation, data association, sensor fusion, fault detection.
Modeling and simulation: linear state space, nonlinear state space, discrete-time state space, matrix-exponential discretization, stochastic state models, stochastic differential equations, hybrid systems, time delays, delay augmentation, state-command structures, integral state augmentation, disturbance estimation, saturation, transfer functions, discrete transfer functions, uncertainty models, block diagrams, signal-flow graphs, simulation, discrete-time simulation, event handling, co-simulation, system identification, nonparametric identification, parametric identification, black-box identification, least-squares identification, recursive least-squares identification, maximum-likelihood identification, subspace identification, identification experiment design, first principles, linearization, canonical forms, similarity transformations, minimum realizations, model reduction.
System analysis: performance, stability, margins, Nyquist, Bode, Nichols, root locus, phase plane, pole-zero plots, Routh-Hurwitz, Jury tests, bilinear-transform stability tests, Nyquist criterion, Bode's relations, fundamental limitations, spectral-density analysis, passivity, sensitivity, steady-state accuracy, z-plane geometry, controllability, observability, nonminimum phase, Lyapunov stability, control Lyapunov functions, control barrier functions, reachability, invariant sets, formal verification.
Core concepts: feedback, feedforward, continuous time, discrete time, frequency domain, Laplace domain, Z-domain, C2D, D2C, positive feedback, feedback with logic, control-system architectures.
First-principles examples: mechanical, electrical, process-control, robotics, aerospace, thrust vector control, computing and network models, biological models, pharmacokinetics, population dynamics, operational amplifiers, atomic-force microscopes, power systems, and precision motion.

Contributing

Contributions are welcome. Useful additions include:

Missing map topics or clearer descriptions.
Canonical textbooks, courses, papers, software libraries, and examples.
Corrections to terminology, equations, or categorization.
Short learning paths for specific domains such as process control, robotics, aerospace, power systems, or autonomous vehicles.

Please keep additions concise and consistent with the map-aligned structure.

License and Attribution

The original README text in this repository is licensed under a Creative Commons Attribution 4.0 International License.

The control theory map graphic is by Engineering Media / Brian Douglas and is included here with attribution. See the original map page for the creator's sharing terms:

Engineering Media map page
Direct PNG
MATLAB YouTube Channel video explanation: Understanding Control Systems

Name		Name	Last commit message	Last commit date
Latest commit History 29 Commits
website		website
Control_Map_ver5.png		Control_Map_ver5.png
README.md		README.md

Folders and files

Latest commit

History

Repository files navigation