How to get intermediate point in Hermite-Simpson collocation method

136 Views Asked by At

I'm reading J. Betts' book "Practical Methods for Optimal Control and Estimation Using Nonlinear Programming" and in the chapter on IVP there is Butcher tableau for Hermite-Simpson method: $$ \begin{array}{c|cccc} 0 & 0 & 0 & 0 \\ \frac{1}{2} & \frac{5}{24} & \frac{1}{3} & -\frac{1}{24} \\ 1 & \frac{1}{6} & \frac{2}{3} & \frac{1}{6} \\ \hline & \frac{1}{6} & \frac{2}{3} & \frac{1}{6}\end{array} $$

General notation used is the following: $$ \begin{array}{c|cccc} \rho_1 & \alpha_{11} & \ldots & \alpha_{1K} \\ \vdots & \vdots & & \vdots \\ \rho_K & \alpha_{K1} & \ldots & \alpha_{KK} \\ \hline & \beta_1 & \ldots & \beta_k \end{array} $$

Also there are formulas: $$ \begin{align*} &\mathbf{y}_{i+1} = \mathbf{y}_i + h_i\sum\limits_{j=1}^K \beta_j \mathbf{f}_{ij} \\ & \text{where } \mathbf{f}_{ij} = \mathbf{f}\left[\left(\mathbf{y}_i + h_i\sum\limits_{l=1}^K \alpha_{jl}\mathbf{f}_{il} \right), (t_i + h_i\rho_j)\right] \end{align*} $$

I don't understand how to get formula for intermediate point $\mathbf{y}_{i+1/2} = \frac{1}{2}(\mathbf{y}_i + \mathbf{y}_{i+1}) + \frac{h_i}{8}(\mathbf{f}_i - \mathbf{f}_{i+1})$ using these formulas above.

So I use the coefficients from the second row and get some ugly formula: $$ \mathbf{y}_{i+1/2} = \mathbf{y}_{i} + h_i \left(\frac{5}{24}\mathbf{f}_{i1} + \frac{1}{3}\mathbf{f}_{i2} - \frac{1}{24}\mathbf{f}_{i3}\right) $$ and I completely stuck, the only thing I managed to see is that $\mathbf{f}_{i1} = \mathbf{f}(\mathbf{y}_i, t_i) = \mathbf{f}_i$. But it doesn't help much. I need help because I'm confused

1

There are 1 best solutions below

0
On BEST ANSWER

Using a little different notation, let's rewrite it in the following form: \begin{align} \mathbf{y}_{i+1/2} &= \mathbf{y}_{i} + h_i\left(\frac{5}{24}\mathbf{f}_{i} + \frac{1}{3}\mathbf{f}_{i+1/2} - \frac{1}{24}\mathbf{f}_{i+1}\right) \\ &= \mathbf{y}_{i} + \frac{h_i}{2}\left(\frac{5}{12}\mathbf{f}_{i} + \frac{2}{3}\mathbf{f}_{i+1/2} - \frac{1}{12}\mathbf{f}_{i+1}\right) \end{align}

Let's now notice that $\frac{5}{12}\mathbf{f}_{i} = \frac{1}{6}\mathbf{f}_{i} + \frac{1}{4}\mathbf{f}_{i}$ and $-\frac{1}{12}\mathbf{f}_{i+1}=\frac{1}{6}\mathbf{f}_{i+1} - \frac{1}{4}\mathbf{f}_{i+1}$. So we get: $$ \mathbf{y}_{i+1/2} = \mathbf{y}_{i} + \frac{h_i}{2}\left(\frac{1}{6}\mathbf{f}_{i} + \frac{2}{3}\mathbf{f}_{i+1/2} + \frac{1}{6}\mathbf{f}_{i+1}\right) + \frac{h_i}{8}\left(\mathbf{f}_i + \mathbf{f}_{i+1}\right) $$

Now let's notice that: $$ \mathbf{y}_{i+1/2} = \frac{\mathbf{y_i}}{2} + \frac{1}{2}\left(\underbrace{\mathbf{y}_i + h_i\left(\frac{1}{6}\mathbf{f}_{i} + \frac{2}{3}\mathbf{f}_{i+1/2} + \frac{1}{6}\mathbf{f}_{i+1}\right)}_{\mathbf{y}_{i+1}}\right) + \frac{h_i}{8}\left(\mathbf{f}_i + \mathbf{f}_{i+1}\right) $$

And finaly $$ \mathbf{y}_{i+1/2} = \frac{1}{2}\left(\mathbf{y}_{i} + \mathbf{y}_{i+1}\right) + \frac{h_i}{8}\left(\mathbf{f}_i + \mathbf{f}_{i+1}\right) $$