一般的线性方程我们可以用最小二乘来解，一般的非线性方程我们可以用LM来解。

这里是线性微分方程组，所以我们采用最小二乘来解。

关键是构造出最小二乘形式，微分可以通过前后数据差分的方法来求。

不过这里还有一个技巧就是如果数据前后帧间隔过大，可以先插值，再对插值后的数据差分如果实际测量数据抖动过大导致插值后差分明显不能反映实际情况，可以先对数据平滑（拟合或是平均）再求差分。(We can use least squares to solve general linear equations, and we can use LM to solve general nonlinear equations.

Here is a system of linear differential equations, so we use least squares to solve.

The key is to construct the least squares form, the differential can be obtained by the method of difference between the front and back data.

However, there is another trick here. If the data is too long before and after the frame interval, you can first interpolate, and then the data difference after interpolation. If the actual measured data jitter is too large, the difference after interpolation obviously can not reflect the actual situation, you can smooth the data first Sum or average) and then find the difference.)