Torque is a very important aspect of motion control.  Torque in a car (electric or combustion powered) is what turns the tires.  Torque control is the ability to control the amount of torque needed based on conditions of the application.

In the car, starting torque requirements can be very high, depending on how fast you want to accelerate.  Running torque, the amount of torque needed to overcome air drag, is very small.   If you want to reach 60 miles an hour in 4 seconds or less, like a Corvette, you will probably need around 450 horsepower.   Cruising on the freeway at 65 miles an hour, constant speed, will probably only require 5 horsepower.

So the power requirements for what appears to be “the same” application can vary wildly, 100 times the power, depending on the circumstances.  That’s where control and regulation come in.   The question being, how fast is the rate of power being dissipated over a small instant in time.  This is the domain of calculus, the first derivative of power over time.  This will determine how fast the control system monitors and updates the value of the power being controlled.  This can also be referred to is as the dynamic rate of the control system, the change in time for the power rate to be measured.

In AC drives, the dynamic response of the drive is a crucial parameter in order to specify the right drive for the application.  Large systems like paper rolls, which can weigh hundreds of pound, have a very slow dynamic rate and a drive for this application should have a dynamic response that is comparable.

Hard disk spindle drives, which have tiny loads and must regulate speed and acceleration based on 2 millisecond seek times, must have extremely fast dynamic response.  The high rate of acceleration requires that torque is regulated in the microsecond range.  Regulating a paper drive with a control designed for hard disks would not only be a waste, but the high response rate in the control would probably lead to instability in the control.

But more complex conditions exist in the real world that must be considered.  What happens when the load is changing?  When you have to palletize beer bottles, every ten cases of beer completes a layer on the pallet.  There are 8 layers to the pallet.  So you start with an empty pallet and end up after 8 identical moves with 3800 pounds of beer.  How do you set the gain?

Either you use an average value equivalent to half the payload weight and live with the results, or you need something that reloads the gain value as the load changes.  Both techniques can be done, both work, but the ideal solution is the second, adaptive gain.  Something that is adaptive, however will require some pretty advanced programming to consider all possible conditions.

And that is the new frontier in robotics.  If robots are to work in human service, they have to be able to operate in a reduced torque mode so that they cannot produce forces that exceed human strength and frailties.  But there are other conditions where the robot’s superior strength can be extremely helpful.  So the current generation of drives will have to incorporate increasingly complex adaptive gain controls in order to make human service robots safe and practical.

Adaptive gain is a discipline that has been talked about for at least a decade in the control community, but it’s been somewhat of a technology looking for an application.  There are the occasional situations where adaptive gain would really be “cool”, but not any widespread applications.  Well, the next great application for adaptive gain will likely be human service robots.  Coming soon to a neighborhood near you.