© Collier Research Systems | Charlottesville, VA

Search
  • Zachary Collier

Why Flexibility and Feedback are Essential for Control

In a recent blog entry, I wrote about the importance of recognizing what is within your control and what is outside of your control. Since then, I have been thinking more about the relationship between planning and control. If you cannot control everything, how can you effectively plan? What exactly does it mean to control something, anyway?


For example, in the engineering sub-discipline of control theory, the goal is to design systems that:

- Achieve the desired outputs

- Avoid wild oscillations

- Settle quickly to steady state values (1)


In non-engineering language, the bullet points can be re-stated:

- Make sure the system does what you want it to do

- Make sure the system doesn’t do what you don’t want it to do

- If something weird is happening, get it to go back to normal as soon as possible


A classic example is a thermostat. Your goal is to keep your house no more than a certain temperature, say 72 degrees. The thermostat has a temperature sensor that reads the current room temperature. A controller compares the desired temperature to the current temperature. If the current temperature is warmer than the desired temperature, it turns the cooling system on, once the current temperature is at or below the desired temperature, it turns the system off. This process of feedback ensures that the room temperature, maintains the desired temperature (first bullet point), doesn’t wildly fluctuate from hot to cold (second bullet point), and in the case of a disturbance (say it was turned off accidentally during a hot day), quickly cools the room to the desired temperature (third bullet point). The process, in other words, controls the temperature in the room.


This sort of control uses feedback and a set of simple if-then rules (e.g., if the room is too warm, turn on the AC). In one of my previous posts, I compared planning to using a map, and hypothesized that good plans have flexibility built into them. For example, my plan might be to drive to Richmond using Interstate 64, but if traffic is stopped due to an unforeseen accident or construction, then I can take some secondary roads to get to my destination. This flexibility uses the same principles as the thermostat: feedback (sensing the current state of the road compared to the desired state) and application of rules (if traffic is stopped, then take the back way; if traffic is smooth, then keep going). In this way we are controlling our commute through a flexible kind of planning. (Note that this type of planning is even better if it is not simply reactive, but also leverages prediction, but that is a discussion for another blog entry).


What are the implications?


Good planning is forward looking and goal-directed. Plans should recognize the uncertainty inherent in the future, and change accordingly when conditions change and new information is gained. This requires one to incorporate feedback mechanisms which enable flexibility. The effective use of feedback is important to effective control. Rather than stating “this is how it needs to be done”, feedback involves stating “this is our goal” and periodically asking “is what we are doing working?”. If things are working, great, if not, there need to be capabilities in place to course correct.


The result is seemingly paradoxical – control is not achieved through rigid, top-down enforcement, but rather through flexibility and feedback.


At Collier Research Systems, we can help you develop strategies and decision making tools to harness uncertainty and help you plan for the future. To learn more, visit (www.collierresearchsystems.com).


References:

(1) Abdelzaher, T., Diao, Y., Hellerstein, J. L., Lu, C., & Zhu, X. (2008). Introduction to control theory and its application to computing systems. In Performance Modeling and Engineering (pp. 185-215). Springer, Boston, MA.