Hi,
I regret that i can not remember the details of how we did this back in the 1980's. It was a long time ago and i had to hand over all my notebooks when i left the company, for obvious reasons.
On the other hand it may depends on the motor being used too, so you probably have to ask the manufacturer. If they can help you, then you are suck with experimenting until you find a workable solution. You would have to be able to apply mechanical loads to the motor shaft too in order to do these kinds of tests.
One thought that comes up is that maybe you can incorporate some feedback. Measuring various parameters might tell you what is best, and that could feed back to the control circuit. For one simple example, you could measure the shaft speed and if it goes too low then increase some parameter, measure current, look for power used, decide what is best. I would think that a minimum power solution would be good.