mb-29 / greedy-identification

Online active identification algorithm for linear dynamical systems.

Greedy online identification of linear dynamic systems

Implementation of our greedy algorithm along with baselines.

Example

from agents import Random, Greedy, Gradient

# ...

def dynamics_step(x, u):
            noise = sigma * np.random.randn(d)
            return A_star@x + B@u + noise

agent_ = Greedy  # or Random, or Gradient

agent = agent_(
    dynamics_step,
    B,
    gamma,
    sigma,
    prior_estimate,
    prior_moments,
    )

estimates = agent.identify(T)