Rise of the Robots: A Theory of Control

Driverless Car from http://www.willappsug.com/2014/01/bmw-introduces-its-autonomous-driverless-car/

With the rise of robotic cars, there is a risk that the following may be a very dated example within only 20 years. 

But consider this. When you are manually driving your automobile, you press on the accelerator pedal to get more speed. And because of friction and drag, releasing the pressure on the pedal will decrease speed. Perhaps you are new to driving, so you are getting all your speed information from the dial on your dashboard. 

Most cars nowadays have a little robot that does this for you which is called a “speed control”.

Whether you do it or the robot does, there is a feedback loop that decides the movement of the accelerator to increase or decrease speed. If the feedback loop is you, when you see the speedometer rise above the speed you want to go, you release the pressure on the accelerometer. If you are going too slow, you push a bit more. Your feedback can be mild or vigorous. If it is too mild, you take longer to approach the speed you want from either too slow or too fast. If it is too vigorous, you may overshoot the speed you want and have to correct again. Probably, if you are too vigorous, you will decide to dial back your enthusiasm to a milder input. Or perhaps you know this car so well, you already know what input to give. 

Every system has a “natural frequency” that describes the worst place to set the feedback. For instance, some young drivers lurch the car fast and slow because they react just at the wrong moment to hit the pedal, or the lurching of the car interacts with their foot on the accelerator pedal. Aircraft will crash if the control system excites the natural frequency of the system (the aircraft). The man-in-the-loop, the pilot, is usually the unstable element. The great inventor, Nikola Tesla, thought he could shake a city apart by finding its natural frequency and providing just a little push at the right frequency. There is some truth to that: The Tacoma Narrows bridge collapsed due to its natural frequency. 

Tacoma Narrows gif from http://giphy.com/gifs/cinemagraphs-bridge-tacoma-duatwzNErHFKw

Think of all of this in a general way. A system (car) with an input (accelerator pedal) and output (speedometer dial). A feedback loop (you) corrects to get the system to the stable state (correct speed). This same mechanism appears everywhere in our human inventions, but also everywhere in nature.

In man made devices, there is a “gain knob” somewhere in the feedback that allows the response to be mild or vigorous. The gain can be dependent on the past so that too vigorous of a response can be decreased on the next try, or visa versa. Models of the system you are controlling can be included in the controller to help make the controls right the first time. Other man made examples are the thermostat in your house which watches the temperature and turns on or off the furnace, steering nozzles on the end of your spaceship which try to keep you on course, or the way you focus your camera. I am sure you can think of dozens more.

Biological creatures have feedback loops. For instance, eating sugar can trigger the release of insulin to control blood sugar levels. Ecosystems can have control loops such as the cycle of predator and prey which keeps both at balanced levels. Or the cycle of disease where the more aggressive a disease, the more likely the hosts will not spread it. Everyone has run into people who seem to be “off their meds” with an out-of-control brain. 

Climate scientists study global control loops and worry that a tipping point will be reached where the control loop can no longer keep the planet’s climate stable. Will it start out like a teenager lurching along in his car, ending in a crash?

Inherent in all this is the question of whether a system can be stable, no matter how much effort you put in to feedback loops. Perhaps the system cannot be observed well enough to control? It would certainly be harder to control an automobile’s speed if it did not have a speedometer. We like to think we can control our economy, but what can we observe for sure and are there sufficient levers to pull to make a difference?

Some people are confused about how to control things. They often seem to want to break the glass and grab the speedometer needle in order to slow the car. When that doesn’t seem to work, they simply try harder to push the needle. Others don’t know how to get control of the accelerometer pedal to speed up or slow down. For instance, CO2 increasing in the atmosphere seems to have a lot to do with climate change. But what if it is highly unlikely you can reduce the levels fast enough given all the emerging nations that desire the same growth curve as the first world? Is there an alternate control loop? Perhaps have airliners emit a gas that will counter the effects of CO2? Will your alternative work about as well as hitting the brakes and accelerator at the same time (which is not very well at all)? Can we financially afford the new control loop or is there one that works about as well but is much cheaper?

The answer lies in systems engineering and systems management principles. These principles were developed in the mid-1950’s when the nation had to create sea launched and intercontinental ballistic missiles and get them deployed as fast as possible. These highly complex systems had to be 100% reliable and safe despite the thousands of moving parts. Hardware, software, and facilities had to be built. People had to be trained. Foolproof procedures had to be created. Supply chains had to be maintained. And it all had to integrate together. 

Some important questions these disciplines help you answer:  What is the system you are trying to control? Can you model it using math and computers? What is its natural frequency? What is the cost to observe the output and control the input? Will the controller always succeed in making the system stable? How might it fail? 

Within Systems Engineering and Systems Management there are plenty of science, engineering, technology, and management tools to deal with complex systems that must be controlled. The designer (hopefully!) has the background and knows where to find them.

Anyway, this is a very quick and dirty overview of something called "control theory". It is the kind of engineering that I studied in school, eventually getting a bachelor’s and master’s degree. The BS was in Astronautical Engineering and the MS was in Mechanical Engineering. I also have an MS in systems management. What drew me to these disciplines?

Everyone wants to control their lives. Some are rabid about it. Others, like the homeless, feel a great loss of control. Financially successful people have created life disciplines to keep them on track and in control. 

I noticed at a very early age how some things and some people could swing wildly out of control. How could a Dad be there one day, and dead the next? Why did my WWII Navy Vet Uncle act so upset and angry seemingly at random? Could my neighbor’s kids go more than a day without getting beat by their Dad? Nervousness, depression, and other moods seemed to always be in control of some people while others charted a stable course through life.  There certainly seemed to be a tipping point for some people. 

Within my own self I could find the levers to handle my anger issues. But was the gain set too high or too low? If I feel sad too much of the time, then I could always adjust the “help other people” lever. Will the feedback loop act in minutes or weeks? Is there a natural frequency to human behavior that helps or thwarts control? These were questions I was asking myself even as a pre-teen.

And another question: "Can people, in general, adequately control themselves?" A short drive to the store will answer this for you.

Well, frankly, if you can’t drive your car better than you do now, we’ll just have to have robots do it for you.