Eric Brasseur Home    |    Links    |    Contact    |   

Electric motors for artificial muscles

There is one domain in which we are largely behind what novel writers "promised": cybernetics. One explanation of that fact may be that compact, powerful and versatile artificial muscles aren't available. This article tries to explain the problem and give some possible solutions. Basic formulas and technical data are available in the appendixes.

Two motors in parallel or serial
Changing the magnetic field of the stator
Increase the electric voltage at high speeds



Today, electric motors from all sizes, as well as modular gear boxes, are available for a few $.

Imagine you want to make a robot arm of a half meter long, capable of lifting a weight of 5 kg a few centimetres up in a few seconds. The movement being rather slow, you can use a huge gear box, and a little motor with a diameter of 3 cm, costing about $ 2, and consuming only 10 W.

However, if you try it out, you will observe one severe drawback with conventional little electric motors: they only give their full power when turning around one specific (ideal) speed. And their absolute maximum speed, with no load, is only twice that speed. So, your robot arm, designed to slowly lift heavy weights, will always make slow movements. Even with no weight to lift at all.

You can't turn the problem by designing the robot arm for high speeds (with no weight), and hope it will just slow down when lifting heavy things: when the motor slows down beneath it's ideal speed, the amount of energy delivered decreases faster than the speed. So, if an electric motor can't lift a weight when turning at its ideal speed (which, as we said, is one half of it's maximum possible speed), it won't lift it at any speed at all. It will stop, or even turn back.

A robot arm using a conventional little electric motor and a gear box will always move at approximately the same speed, whatever the weight to lift. If the weight is too heavy, the arm will stop completely.

For standardized industrial applications, this may not be a crucial problem: robot arms are made for well defined purposes, and are optimized for them. Moreover, robot arms are generally fixed, and their weight, price and power consumption are of no great importance. They can use over-dimensioned motors, capable of fast and powerful movements. What matters most is the precision with which they can make those movements. (For such purposes, stepping motors are a good example of heavy, big, expensive, energy consuming and low-efficiency, but accurate, devices.)

But for universal, autonomous and lightweight cybernetic systems, it is a disaster. They must be capable of both lifting slowly heavy weights, and move fast with light weights.

Certain solutions, like miniaturised mechanical or hydraulic gear changes, are possible. But they will be expensive, fragile, and delicate to control.

It would probably be better to use more classical hydraulic solutions, like little pistons and miniaturised floodgates. This will also be rather expensive, and maybe dangerous, because of the risk of one floodgate remaining open, and making an arm or finger "crunch" what it holds.

It is also possible to make some sort of rustic gear changes by using all sorts of systems like loosely hanging belts which are stretched when low speeds and high forces are necessary, or belts which are normally rigid but become springs when one pulls too hard on them, or pushing levers, or cogwheels with only a part of their tooth... But those sorts of solutions are often voluminous, unreliable, slow in their reactions, making sudden an uncontrolled movements, and as expensive to manufacture as they had been cheap to imagine.

Here are three more "realistic" solutions, using electric motors:


Two motors in parallel or serial

This is a solution inspired by mother nature: make the arm move by two different motors. One motor is made for slow and powerful move ments, the other for fast and less powerful movements.

Medical research has shown that our muscles are made of two types of fibres: "fast" fibres, and "slow" fibres. The first are for quick move ments, the later for slow and strong movements. Whatever movement you make, a part of your muscular fibres is useless. Even worse: it's extra weight, and disturbs the working fibres by it's presence. But it works.

Two motors in parallel

During slow movements, the motor for fast movements should be cut off. Because it will turn very slowly, and would transform a lot of electric energy into heat by Coulomb effect. During fast movements, the motor for slow movements should be disconnected. Because it will turn very fast, and behave like an electric generator of higher voltage than the normal power source. It could burn the rest of the system, or act as an electromagnetic brake.

The most important drawback with this solution is that at high speeds the slow motor can act as an inertia wheel, and slow down the changes of speed. This can be partially compensated by using burst modes on the slow motor.

Putting two independent motors in parallel will make mechanical problems arise: strains, friction between cogwheels... (Just try to make a cogwheel system turn by forcing on the "slow" end.) The solution (still using conventional systems), is to put the motors closer to each other. They should only be separated by one or two cogwheels, and have the rest of the gear box in common:

Both motors can even be put on the same axis, provided they are different. One should be slow and strong, the other fast and weaker. (If they use equivalent technologies, the slow one will be flat with a large diameter, the fast one will be long and thin.) (A simpler way to make such different motors, is by using two the same motors, and weaken the permanent magnet around one of them.) This system should not be over-estimated. Especially because the permanent magnets usually used are already very strong, and because little motors turn at a speed not far from their mechanical maximum speed. So some sort of gear change will probably be necessary: the "slow" motor should only be mechanically connected when slow movements are done. This could be achieved just by moving a single little cogwheel a few millimetres aside. (The volume of a whole encyclopedia wouldn't be enough to describe all the different ways to make that system, not necessarily based on a moving cogwheel.)

(Note that this solution can also be used for turbines, which do also have the problem of giving their full power only when turning in a small window of speed.)

Two motors in serial

The best way to describe this system is probably the following:

(This diagrammatic conception does not take the bending efforts in consideration.) Both motors must have a brake, so they can be blocked when not in use.

Such brakes are easy to make: a little electromagnet just has to stick itself to a cogwheel close to the motor. Like what happens in relays. (Of course, once a motor is blocked by its brake, it shouldn't be fed any more with electric power.)

The advantage of this system is certainly that the speeds achieved by the two (or more) motors can be totally different. Moreover, no problem arises from interactions between the two gear boxes, like for motors in parallel.

Provided both motors are different, they can be put in serial before their common gear box. The rotor of the slow motor can be the stator of the fast one.

(Again, this trick can be used for turbines.)


Changing the magnetic field of the stator

If you increase the magnetic force of the stator of an electric motor, you will reduce the maximum speed of the motor, as well as its ideal speed. However, the power delivered at this new ideal speed will remain the same as before.

The increase of the magnetic field can be obtained by a static electromagnet put above little permanent magnets. When you want an increase in the mechanical power given by the motor at low speed, you simply switch on the electromagnet.

Note that the power of the stator can not be increased or decreased indefinitely: too power ful, it will override the magnetic field of the rotor, and block it. Too weak, it will be overridden by the magnetic field of the rotor, and be of no use.

This system can be self-regulating: at high speeds the electrix counter-force lowers the current in the stator. This system can also take profit of the trick illustrated in the next chapter.


Increase the electric voltage at high speeds

When a motor is turning slowly, increasing exaggerately the electric voltage at its terminals is very risky: the current passing trough the coils will become too strong, and so the heat produced. The motor will burn, or at least have a shorter life.

But at high speeds, an increase in electric voltage will simply compensate the counter- force exerted by the stator on the coils of the rotor. That counter-force which was indeed responsible for the drop in the power delivered by the motor, when turning above a certain speed.

This augmentation of electric voltage with the speed can easily and reliably be obtained simply by imposing a constant current into the motor. Probably best by using high voltage power sources and low-volume, approximate, switching power supplies. The highest voltage should be the one that makes the motor turn at its mechanically maximum speed. (The best results will be given by a system making the voltage drop when the motor approaches speed where resonances and frictions appear.)

This solution is the easiest to implement on existing servomechanisms. An amateur could do it simply by adding a little electronic circuit to an existing servo for model kits.

(It is probable that many little electric motors, whose official operating voltage is given so that they won't burn when blocked, will be able, with this system, to deliver much more power. The current forced into them should simply be the current that flows trough them when they are blocked and connected to their normal power source.)



A lot of variants of the four solutions given above are possible. They are not mutually excluding, and solutions exist which are in- between them.

Let's hope the manufacturers will soon provide us with integrated solutions; strong and lightweight servomechanisms capable of large varieties of speeds, high efficiencys, the possibility of easy electronic command, and remote reading of their position and/or speed.

Electronic auto-control of the servomechanisms can easily be achieved by modern microcontrollers. (The most little microcontrol ler of the PIC 16C5X family of Microchip costs $ 2, is hosted in a thumbnail-big 18-lead SMD, is a 2 MIPS RISC machine with ROM and RAM, and consumes 2 mA.)

One feature those servomechanisms should have, and humans don't have: brakes. A considerable amount of energy can be saved just by being able to stuck a muscle in a given position.

Apart from the now very mediatic fuzzy logic, constructors of servomechanisms should be aware of the achievements of M. Rodney Brooks for the realisation of light weight, fast and autonomous systems adapted to the real world. It offers instinctive artificial intelligence at the speed of a few electronic computations.

Although they cannot be found in the common stores, there are three types of electric motors that should be mentioned:

Eric Brasseur  /  Didier Bizzarri  -  1994