====== Planning and control ======

===== Model predictive control =====

  * Backprop through kinematic equation
  * Minimization wrt the latent

==== State transition equations ====

$\dot{x}=f(x,u)$

$u$ control (latent variable)
$x$ state

$\frac{dx(t)}{dt} = f(x(t), u(t))$

temporal function $t$