Calculus of Variations

Optimizing a Whole Path

Ordinary optimization picks the best number: the profit-maximizing price, the cost-minimizing batch size. The calculus of variations picks the best function — an entire path through time.

The object being scored is a functional: a rule that eats a whole curve and returns one number.

$$J[x] = \int_0^T F\big(t,\, x(t),\, \dot{x}(t)\big)\,dt$$

Examples of $J$: total cost of a production ramp-up plan; lifetime utility of a consumption path; total fuel along a trajectory. Change the path anywhere, and the score changes.

The classical warm-up: which curve between two points has the least length? Of all the infinitely many candidates, the straight line wins — and proving that (rather than assuming it) is the field's founding exercise.

The Euler–Lagrange idea. At an optimal number, the derivative is zero. At an optimal path, no small deformation — no 'variation' — of the path can improve the score. Pushing that condition through the integral yields the Euler–Lagrange equation:

$$\frac{\partial F}{\partial x} - \frac{d}{dt}\frac{\partial F}{\partial \dot{x}} = 0$$

A differential equation the optimal path must satisfy at every instant — local perfection everywhere, forced by global optimality.

Economic classics built exactly this way: the Ramsey model (how a nation should split output between consumption and investment over decades), optimal advertising spend over a product's life cycle, extracting a finite resource so profit is highest across the whole horizon, not just today.

The mental upgrade this lesson installs: stop asking what is the best level? and start asking what is the best trajectory? — because most business decisions are movies, not photographs.

Tip: Keep the object hierarchy straight: a function eats a number and returns a number; a functional eats an entire curve and returns a number. Optimization over functionals is choosing the best whole path, not the best single value.

Common pitfall: Solving a path problem by optimizing each instant separately. The Euler–Lagrange logic exists because the best path balances across time — sprinting production today changes tomorrow's costs, and only the whole-path view sees it.