Collocation method

Content sourced from Wikipedia, licensed under CC BY-SA 3.0.

Collocation method: a simple way to solve complex equations

What is it?
A collocation method is a way to approximate solutions to ordinary and partial differential equations by using a short polynomial and forcing the equation to hold at a few chosen points, called collocation points.

How it works (for y′(t) = f(t, y(t)), with y(t0) = y0)
- Solve on a small interval [t0, t0 + h].
- Pick n collocation points ck in the interval [0, 1], and set tk = t0 + ck h.
- Approximate y(t) by a polynomial p of degree n. It satisfies the initial value p(t0) = y0 and the differential equation at the collocation points: p′(tk) = f(tk, p(tk)).
- This gives n + 1 conditions, which matches the n + 1 coefficients of the degree-n polynomial. The method is implicit (it involves y at unknown future points).

Relation to Runge–Kutta
All collocation methods are implicit Runge–Kutta methods. The collocation points ck appear in the Runge–Kutta coefficients. Not every implicit Runge–Kutta method is a collocation method, but collocation methods are a special family of them.

A simple example: the trapezoidal rule
- If we pick two collocation points at the ends, c1 = 0 and c2 = 1, the method uses a quadratic p and leads to:
y1 = y0 + (h/2) [ f(t0, y0) + f(t0 + h, y1) ],
where y1 ≈ y(t0 + h).
- This is the trapezoidal rule, which can also be seen as approximating the integral form y(t) = y0 + ∫ from t0 to t f(τ, y(τ)) dτ with the trapezoidal rule.

Gauss–Legendre collocation
- Uses Gauss–Legendre points as collocation points. With s points, it has order 2s and is A-stable (good stability for many problems).

Direct vs orthogonal collocation
- Direct collocation uses piecewise polynomials in time (like the trapezoidal rule or higher-degree pieces) to approximate the solution.
- Orthogonal collocation uses a basis of orthogonal polynomials (for example Legendre polynomials) to build the approximation.

Why it matters in practice
Collocation turns a time-evolution problem into an optimization or algebra problem, which can handle complex dynamics and constraints. For example, in engineering and racing simulations, it can be used to optimize schedules and controls (like throttle, brakes, and steering) while respecting the physics at every point in time. This approach enables finding the fastest or best-performing solutions under many constraints (energy, temperature, tires, etc.) in a single, coherent calculation.