Intro to modeling

In this course, a model is a mathematical stand-in for a physical system that captures the relationships we care about, especially how the system’s outputs evolve over time in response to inputs. In control, we don’t model things just for the sake of having equations: we model so we can predict behavior, analyze stability and performance, and design controllers that work reliably on the real system.

A useful way to think about a model is as a structured prediction tool:

• Inputs: what we can command or what acts on the system (forces, torques, voltages, flows, disturbances).

• Outputs: what we measure or care about (position, temperature, speed, pressure, concentration).

• Dynamics: how the present influences the future (memory), typically written as differential equations (and sometimes algebraic constraints).

Models are purposeful simplifications¶

No model is “the truth.” Every model throws away details. The goal is to keep enough detail to answer the question we’re asking, no more, no less. Too simple, and we get poor predictions and misleading controller designs. Too complex, and the model becomes hard to analyze, hard to simulate, and hard to use for design, with little practical benefit.

So modeling is really the art of making assumptions that are appropriate for the application:

• What effects matter at the time scale of interest?

• What can be neglected without changing the behaviors we care about?

• What operating conditions (small motions? near equilibrium? low speeds?) will the model be valid for?

The modeling workflow we’ll use¶

In the next several lectures we’ll practice a consistent approach for turning physical systems (mechanical, electrical, hydraulic, electromechanical,...) into equations we can analyze and control.

A typical workflow looks like this:

1. Choose variables: define inputs, outputs, and internal variables (often positions/velocities, currents/voltages, pressures/flows, etc.).

2. Draw the physics: free-body diagrams and element laws (springs, dampers, masses, motors, valves,...).

3. Write governing equations: conservation laws / Newton’s laws, extract differential equations, sometimes with algebraic constraints.

4. Simplify intentionally: apply assumptions to get a model that is tractable and fit for purpose.

5. Check the model: do the units make sense, does it match limiting cases, does it predict qualitatively correct behavior?

A quick example of “fit for purpose”¶

Consider a pendulum. A very detailed model might include aerodynamic drag, bearing friction, flexing of the rod, and so on. But depending on the question, we might intentionally ignore many of those effects. For instance, if we care about small oscillations over a short time horizon, we might assume a rigid rod, negligible drag, frictionless hinge, constant gravity, and small angles (so $\sin\theta \approx \theta$). Those assumptions yield a simpler model that is easy to analyze and good enough for control design in that regime.