Control type for constant-speed actuator

I would use a closed-loop (feedback) speed control system with a speed sensor (tachogenerator or encoder) and a PI controller.

 ωref +    e  +----+  +-----+ Va +-------+  ω
 ---->(Σ)---->| PI |->|Amp/ |--->| Motor |-+-->
      ^ -     +----+  |drive|    | +load | |
      |               +-----+    +-------+ |
      |    +------------+            ^ T_L |
      +----| Tacho  Kg  |<-----------------+
           +------------+

Why closed loop

1. Load disturbance rejection: the load torque $T_L$ changes the speed. In closed loop the speed change is divided by $(1+K)$, where $K$ is the loop gain; with the integral term, the steady-state speed error for a step change in load becomes zero.
2. Insensitive to parameter changes: armature resistance rises with heating and the supply voltage varies; the sensitivity of speed to these is $1/(1+GH)$ instead of 1.
3. Faster response: the motor time constant $\tau_m$ becomes $\tau_m/(1+K)$, so the speed recovers quickly after a disturbance.
4. Accurate tracking of the set speed, needed in conveyors, paper mills, fans and pumps.

Why not open loop

An open-loop drive applies a fixed voltage. Its speed is $\omega = \dfrac{V - I_aR_a}{K_b}$, so it falls whenever the load (and hence $I_a$) increases, and it cannot correct the drop.

Why PI

Proportional control alone leaves a steady-state error $1/(1+K_p)$; the integral action removes it. A derivative term is usually not needed for speed loops and it amplifies tacho noise. The gains are chosen to give a fast response with small overshoot.