I must admit that I did not read your OP correctly and missed the bit about "maintain the rpm under load".
You need to determine what sort of percentage variant you will accept, and what cost you are happy to accept. I really feel you need to specify some numbers before you ask again. When you do come up with a spec please post here as well.
I'm just messing about actually with two old motors, which could be used to drive various small things I can think of. I'm not sure seeking speed regulation is that worth it, well, it depends on whether it's expensive or really difficult to achieve. I think it might be made somewhat more problematical because I don't have a Permanent Magnet motor.
My input supply range is up to 20V DC at up to 5A. The motor is 12V, series wound, and has 11 commutator segments. It's a Smiths motor that was once on an blower for an old car. You know, the old squirell-cage type of blower.
I've just found this to mull over:
And I'm searching application notes and other useful web pages.
I think definitely "speed controller" is a key Google search term.
I'm in the middle of a similar project, using 12v 'seat motors' out of a vehicle with adjustable driver and passenger seats. The thought being that if the motor can drag a 20 stone person about it should have no difficulty in doing what I want.
As to the controls, luckily I have a friend who is an electronics whiz and he has suggested the control layout.
Basically, as I understand it, to have a smooth control of a DC motor from low to high revs a Pulse Wave Modulator is needed. See the link to eBay below which shows the control I am using.
Most transformer/rectifiers don't give a 'smooth' output, the voltage can fluctuate quite dramatically and will often produce a voltage a lot higher than stated on the label. For the best results you need a 'smoothed' power supply and although it is possible to build capacitors into the circuit it would be easier to use a car battery and keep it charged with one of the intelligent battery chargers. There is also the argument that if you are driving something which resists being moved by the motor then you don't need a smooth supply as this 'resistance' will do the job for you. But don't overload the PWM!
If you do go down the PWM route make sure you include in the circuit some fast acting fuses to protect the PWM.
I think I should correct some misunderstandings, whether real or imagined I'm not sure.
A PWM circuit just switches on and off very fast, and adjusts the length of the ON and OFF pulses according to some input, often a static voltage from a potentiometer.
This is useful as a sort of "partly-on" electronic switch, where using eg a variable resistor would waste power (dissipated in the resistor), as modern power FETs can switch on and off very quickly.
Normal simple voltage controlled PWM's will vary the ON-ness ratio according to a static voltage, derived from the position of a potentiometer - but this will *not* keep a motor at a specified speed when under varied load, more circuitry is required to increase the ON pulse width when the motor begins to slow down under load.
A speed controller needs more, it needs to have some method of measuring the speed  - then compare that measured speed to the desired speed - then create a difference input to the PWM.
If it does that fast enough, which if should be able to do, then the speed of the motor shouldn't vary much whether it's unloaded or loaded at full torque.
My point is, a speed controller will have this circuitry included - a simple PWN won't. Others have said this, but I'm unsure who understands what here.
 this is easy when a permanent magnet motor is used, as the transitions between segments on the rotor, with changes in the flow of the current, are easy to detect electronically
It is simple to measure the back EMF generated by the motor when the "switch" is off. this gives a linear indication of speed and has been used for the last three decades or so by good controllers.
Conversely, a decent shunt wound, brushless or permanent magnet motor has fairly good speed regulation anyway. Series wound motors should be rewound with shunt field coils if any sort of speed stability is desired.
As this is and engineering rather than electronics group, how about building a centrifugal governor as found on old steam engines? Let the moving arms adjust a potentiometer and you don't need any electronics at all.