Controller/feedback loop

[jkoendev]

for feedback control

[jkoendev]

For a feedback control loop we need

Simulation model (environment)

• with fixed step integrator
• step method to advanced simulation environment: observation = env.step(controls)
• might include some noise

Controller model

• might be different to the simulation model (no noise, simplified model for computational speed)

Controller:

• calculate feedback from starting state controls = solver.feedback(ig, current_state)
• realtime iteration for fast feedback
• use fast/special solver (https://github.com/acados/acados, interface to acados is currently in development)

Optional:

• Estimator/Kalman filter to get current_state (of the controller model) from observation, if no estimator one would assume something like state = eye * observation