synapsys.algorithms

Control algorithm implementations ready for use in agents or standalone scripts.

PID

Discrete-time PID controller with output saturation and back-calculation anti-windup.

Constructor

PID(
    Kp: float,
    Ki: float = 0.0,
    Kd: float = 0.0,
    dt: float = 0.01,
    u_min: float = -inf,
    u_max: float = inf,
)

Parameter	Description
Kp	Proportional gain
Ki	Integral gain
Kd	Derivative gain
dt	Sample time in seconds
u_min, u_max	Output saturation limits

Methods

Method	Description
compute(setpoint, measurement) -> float	Computes the control signal for one timestep
reset()	Clears the integrator and previous error

lqr

Solves the infinite-horizon Linear Quadratic Regulator problem.

lqr(A, B, Q, R) -> tuple[np.ndarray, np.ndarray]

Parameter	Description
A	System matrix $(n \times n)$
B	Input matrix $(n \times m)$
Q	State cost matrix $(n \times n)$, positive semi-definite
R	Control cost matrix $(m \times m)$, positive definite

Returns: (K, P) where K is the optimal gain and P is the Riccati solution.

Source

See synapsys/algorithms/ on GitHub.