Intelligent Systems And Their Societies Walter Fritz

Air Cylinder Control


The use of air cylinders to produce the movement of arms and legs has great advantages and also great disadvantages. Below we explain some of them.


Difficulties with the control of air cylinders

Air supply to the (double action) cylinders is by on-off electro-valves. The duration of an impulse is for an instant (1/15 of a second, 65 milliseconds). There are 3 types of difficulties.

For a typical double action air cylinder with 80 psi air pressure, 35 mm diameter and 105 mm length, the delay between giving an impulse and noting its "effect" is 4 to 5 instants. The "effect" is a change in position of having an impulse compared with no impulse. This delay means that when giving an impulse based on present position and velocity, the effect, 4 instants from now, may not be what we want. We do not wish to improve the present position, which is impossible, due to the time delay. We need to improve the future position. So the decision of which impulse to use has to be based on the position that will exist 4 instants from now.

At the start position the arm hangs down. Gravity has no effect. At the end position the arm is horizontal. The cylinder has to push against gravity (the weight of the arm) and an impulse has much less effect.

When the arm is standing still, we have static friction, which is high. When the arm is moving we have dynamic friction, which is low, here the impulse has much more effect.

Due to the above it is very difficult to write a movement control program, based on physics and mathematics.


The present method of control

To avoid these difficulties, a learning program was written, based on concepts and response rules. For each position sensor (each encoder), the input into the program is the objective position and the objective velocity for getting there. All are expressed as concepts.
The output of the program is the impulse to be made for each cylinder.
The program learns by experience what the effect of an impulse is, in a given situation.

The parts of the response rule are:

With these rules, the program predicts the future position and velocity for each alternative of impulse. Then it selects the best alternative.


The program



Here is a graph of the resulting sideward movement of the left arm of Robby. All other cylinders are inactive. The hand is carrying no object.

The vertical scale indicates the position of the arm. These position numbers is what the angular encoders report. For instance, a move from 11 to 12 corresponds to an angle of 3 degrees.

The horizontal scale is the time in instants. The instant is the time for a cycle of the program and is 65 milliseconds (0.065 sec). So 15 instants are about a second.

At the start, the arm is at the angular position of 12 and is given a command to go to the objective, which is position 25.

At the start of the move we can see a 4 instant delay until motion starts. This delay exists always, even during the movement.

For three consecutive runs, we see that the precision of movement is within 1 angular position number, or the accuracy of the encoder. The objective is reached within .75 seconds.


Comparison with electric motors
A good alternative to air cylinders appears to be the use of DC electric motors with planetary gear head and encoder build in.

The advantage of electric motors is that there is no interval between increasing current and the resulting increase in torque or velocity. So control is easier. Also motors and control can be bought readymade, whereas the required lightweight air cylinders (and air compressor) are not on the market.

The disadvantage is the much higher price and the much greater weight at arms and legs.


Possibly the above may help you in your design decisions for your robot.


For continuous reading, like a book - continue here.
Jump back to e-book Contents / Robby document / the top of this document.

Last Edited 6 Mar. 06 / Walter Fritz
Copyright © New Horizons Press