Single Inverted Pendulum LQR Control Simulation with Implementation Details

Single Inverted Pendulum LQR Control Simulation

The single inverted pendulum is a classic experimental object in control theory, frequently used to validate the effectiveness of various control algorithms. The core challenge involves maintaining the pendulum in an upright balanced position by adjusting the driving force at the base. Since the system is inherently nonlinear and unstable, designing an appropriate controller is essential for achieving stable control. In MATLAB implementation, this typically involves creating state-space models using functions like ss() and defining system matrices that represent the pendulum dynamics.

LQR (Linear Quadratic Regulator) is an optimal control method based on state feedback, suitable for linear systems or linearized systems. By minimizing a quadratic performance index that incorporates state errors and control costs, LQR computes an optimal feedback gain matrix to achieve desired system stability. For inverted pendulum systems, the common approach involves linearizing the nonlinear model first, then applying LQR controller design around the equilibrium point. The MATLAB function lqr() can be used to calculate the optimal gain matrix by specifying the weight matrices Q and R, which balance state regulation and control effort.

In practical systems, some state variables (such as pendulum angular velocity) may not be directly measurable, requiring the design of state observers to estimate these unmeasurable states. Reduced-order state observers aim to reconstruct partially unobservable states using measurable outputs, thereby reducing computational complexity. The implementation often involves partitioning the state matrix and designing observers for only the unmeasured states using functions like obsv() to check observability.

Two design methods for reduced-order observers:

Pole Placement-Based Reduced-Order Observer: This method ensures rapid convergence of estimation errors by selecting appropriate observer poles. The design focuses only on unmeasurable states, reducing computational burden, making it suitable for applications with high real-time requirements. In MATLAB, this can be implemented using place() or acker() functions for pole assignment in the observer dynamics.

Luenberger Observer-Based Reduced-Order Design: The Luenberger observer represents a standard form of state observer, with the reduced-order version estimating partial states. The core concept utilizes system outputs and known dynamic models to construct an estimator structure that ensures stable estimation error dynamics. The implementation typically involves solving matrix equations to obtain observer gain matrices that satisfy stability conditions.

Case analysis typically involves system modeling, linearization, LQR controller design, observer parameter selection, and final simulation verification. In tools like MATLAB/Simulink, by constructing the pendulum's dynamic model and integrating LQR controllers with reduced-order observers, system responses can be visually observed to validate control strategies. The Simulink implementation would include subsystems for plant dynamics, controller blocks, and observer blocks with appropriate signal connections.

Simulation results typically demonstrate pendulum stabilization angle, cart position regulation capability, and controller robustness under different initial conditions. By adjusting LQR weight matrices and observer poles, system dynamic response can be optimized to achieve better performance metrics. The tuning process often involves iterative simulation runs using sim() command and analyzing response characteristics with plotting functions like plot().