# How to obtain the state space representation of a computed torque control system?

Assuming I have a system:

$$a \, \ddot{x}_1 + b \, \ddot{x}_2 + c \,\dot{x_1} + d \,\dot{x_2} + e = u_1 \\ f \, \ddot{x}_1 + g \, \ddot{x}_2 + h \,\dot{x_1} + i \,\dot{x_2} + j = u_2$$

How would I write this system in state space representation: $$\dot{x} = Ax+Bu$$ $$y = Cx + Du$$

Usually, I would isolate for $$\ddot{x_1}$$ or $$\ddot{x_2}$$, but in this case, they are functions of each other. Would the only way be to rewrite the dynamics in matrix form after setting new state variables?

• The fact they are coupled makes no difference. You can use your standard methodology. :), set one for x2 and x1 and then solve. Commented Dec 20, 2020 at 9:07

First, isolate the second order from the other terms:

$$a \, \ddot{x}_1 + b \, \ddot{x}_2 =- c \,\dot{x_1} - d \,\dot{x_2} - e + u_1 \\ f \, \ddot{x}_1 + g \, \ddot{x}_2 = - h \,\dot{x_1} - i \,\dot{x_2} - j + u_2$$

Then, put it in a matrix form:

$$\left[\begin{matrix}a&b\\f&g\end{matrix}\right]\left[\begin{matrix}\ddot{x_1}\\\ddot{x_2}\end{matrix}\right]=\left[\begin{matrix}- c \,\dot{x_1} - d \,\dot{x_2} - e + u_1\\- h \,\dot{x_1} - i \,\dot{x_2} - j + u_2\end{matrix}\right].$$

Supposing $$\left[\begin{smallmatrix}a&b\\f&g\end{smallmatrix}\right]$$ is invertible, we can left multiply both sides by $$\left[\begin{smallmatrix}a&b\\f&g\end{smallmatrix}\right]^{-1}$$:

$$\left[\begin{matrix}a&b\\f&g\end{matrix}\right]^{-1}\left[\begin{matrix}a&b\\f&g\end{matrix}\right]\left[\begin{matrix}\ddot{x_1}\\\ddot{x_2}\end{matrix}\right]=\left[\begin{matrix}a&b\\f&g\end{matrix}\right]^{-1}\left[\begin{matrix}- c \,\dot{x_1} - d \,\dot{x_2} - e + u_1\\- h \,\dot{x_1} - i \,\dot{x_2} - j + u_2\end{matrix}\right].$$

Thus, we get: $$\left[\begin{matrix}\ddot{x_1}\\\ddot{x_2}\end{matrix}\right] = {1\over ag-bf} \left[\begin{matrix}g&-b\\-f&a\end{matrix}\right]\left[\begin{matrix}- c \,\dot{x_1} - d \,\dot{x_2} - e + u_1\\- h \,\dot{x_1} - i \,\dot{x_2} - j + u_2\end{matrix}\right],$$

$$\left[\begin{matrix}\ddot{x_1}\\\ddot{x_2}\end{matrix}\right] = {1\over ag-bf} \left[\begin{matrix}(bh-gc)\dot{x_1}+(bi-gd)\dot{x_2}+(bj-je)\\ (fc-ah)\dot{x_1}+(fd-ai)\dot{x_2} + (fe-aj)\end{matrix}\right] +{1\over ag-bf} \left[\begin{matrix} g&-b\\ -f&a \end{matrix}\right] \left[\begin{matrix} u_1\\ u_2 \end{matrix}\right].$$

Using auxiliary variables: $$\left[\begin{matrix}\ddot{x_1}\\\ddot{x_2}\end{matrix}\right] = \left[\begin{matrix}\alpha\dot{x_1}+\beta\dot{x_2}+\gamma\\ \delta\dot{x_1}+\epsilon\dot{x_2} + \zeta\end{matrix}\right] + \left[\begin{matrix} \eta&\theta\\ \iota&\kappa \end{matrix}\right] \left[\begin{matrix} u_1\\ u_2 \end{matrix}\right].$$

Here, $$\gamma$$ and $$\zeta$$ can represent either uncontrolled inputs or disturbances.

As uncontrolled inputs: $$\left[\begin{matrix}\ddot{x_1}\\\ddot{x_2}\end{matrix}\right] = \left[\begin{matrix}\alpha\dot{x_1}+\beta\dot{x_2}\\ \delta\dot{x_1}+\epsilon\dot{x_2}\end{matrix}\right] + \left[\begin{matrix} \eta&\theta&\gamma\\ \iota&\kappa&\zeta \end{matrix}\right] \left[\begin{matrix} u_1\\ u_2\\ 1 \end{matrix}\right].$$

As disturbances: $$\left[\begin{matrix}\ddot{x_1}\\\ddot{x_2}\end{matrix}\right] = \left[\begin{matrix}\alpha\dot{x_1}+\beta\dot{x_2}\\ \delta\dot{x_1}+\epsilon\dot{x_2}\end{matrix}\right] + \left[\begin{matrix} \eta&\theta\\ \iota&\kappa \end{matrix}\right] \left[\begin{matrix} u_1\\ u_2 \end{matrix}\right]+ \left[\begin{matrix} \gamma\\ \zeta \end{matrix}\right].$$

Creating the state $$s=[s_1\;s_2\;s_3\;s_4]^T=[x_1\;\dot{x_1}\;x_2\;\dot{x_2}]^T$$:

$$\dot{s} = \dot{\left[\begin{matrix}s_1\\ s_2\\ s_3\\ s_4\end{matrix}\right]} = \left[\begin{matrix} 0&1&0&0\\ 0&\alpha&0&\beta\\ 0&0&0&1\\ 0&\delta&0&\epsilon \end{matrix}\right] \left[\begin{matrix}s_1\\ s_2\\ s_3\\ s_4\end{matrix}\right] + \left[\begin{matrix} 0&0\\ \eta&\theta\\ 0&0\\ \iota&\kappa \end{matrix}\right] \left[\begin{matrix} u_1\\ u_2 \end{matrix}\right] + \left[\begin{matrix} 0\\ \gamma\\ 0\\ \zeta \end{matrix}\right].$$

Depending on your system, you can design a controller that rejects those disturbances; the presence of integrators in the system can help remove steady-state errors.

• Thank you. Is this generally the way computed torque control is done? I am wondering if it would be possible to transform this into a second order state space by representing x1 and x2 in a single vector q = [x1;x2]? Commented Dec 20, 2020 at 17:23
• In youtube.com/watch?v=MV-xBPP3H2k, you can see that a combination of feedfoward and PID can be used to control using the torque as input. Commented Dec 21, 2020 at 16:30

\begin{align} a \, \ddot{x}_1 + b \, \ddot{x}_2 + c \,\dot{x_1} + d \,\dot{x_2} + e &= u_1 \tag{1} \\ f \, \ddot{x}_1 + g \, \ddot{x}_2 + h \,\dot{x_1} + i \,\dot{x_2} + j &= u_2 \tag{2} \end{align} Let's write Eq(1) without $$\ddot{x}_2$$, hence: \begin{align} a \, \ddot{x}_1 + b \, \Big[\frac{u_2-f \, \ddot{x}_1 - h \,\dot{x}_1 - i \,\dot{x}_2 - j}{g}\Big] + c \,\dot{x}_1 + d \,\dot{x}_2 + e &= u_1 \\ \left[a-\frac{bf}{g} \right] \, \ddot{x}_1 + \, \frac{b}{g}u_2 - \frac{bh}{g} \,\dot{x}_1 - \frac{bi}{g} \,\dot{x}_2 - \frac{b}{g}j + c \,\dot{x}_1 + d \,\dot{x}_2 + e &= u_1 \\ % \left[a-\frac{bf}{g} \right] \, \ddot{x}_1 = u_1 - \, \frac{b}{g}u_2 + \frac{bh}{g} \,\dot{x}_1 + \frac{bi}{g} \,\dot{x}_2 + \frac{b}{g}j - c \,\dot{x}_1 - d \,\dot{x}_2 - &e \\ % \ddot{x}_1 = \left[\frac{g}{(ag-bf)} \right] \left[u_1 - \, \frac{b}{g}u_2 + \frac{bh}{g} \,\dot{x}_1 + \frac{bi}{g} \,\dot{x}_2 + \frac{b}{g}j - c \,\dot{x}_1 - d \,\dot{x}_2 - e\right] \tag{3} \end{align} Now we do same procedure which is eliminating $$\ddot{x}_1$$ from Eq(2) hence, \begin{align} f \, \Big[\frac{u_1- b \, \ddot{x}_2 - c \,\dot{x_1} - d \,\dot{x_2} - e}{a}\Big] + g \, \ddot{x}_2 + h \,\dot{x_1} + i \,\dot{x_2} + j &= u_2 \\ % \Big[ g - \frac{fb}{a}\Big] \ddot{x}_2 + \frac{f}{a} u_1 - \frac{fc}{a} \,\dot{x_1} - \frac{fd}{a} \,\dot{x_2} - \frac{fe}{a} + h \,\dot{x_1} + i \,\dot{x_2} + j &= u_2 \\ % \Big[ g - \frac{fb}{a}\Big] \ddot{x}_2 = u_2 - \frac{f}{a} u_1 + \frac{fc}{a} \,\dot{x_1} + \frac{fd}{a} \,\dot{x_2} + \frac{fe}{a} - h \,\dot{x_1} - i \,\dot{x_2} - j \\ % \ddot{x}_2 = \left[\frac{a}{(ag-bf)} \right] \left[ u_2 - \frac{f}{a} u_1 + \frac{fc}{a} \,\dot{x_1} + \frac{fd}{a} \,\dot{x_2} + \frac{fe}{a} - h \,\dot{x_1} - i \,\dot{x_2} - j \right] \tag{4} \end{align} Now let $$y_1=x_1, y_2=\dot{x}_1, y_3=x_2, y_4=\dot{x}_2$$, hence: \begin{align} \dot{y}_1 &= y_2 \\ \dot{y}_2 &= \left[\frac{g}{(ag-bf)} \right] \left[u_1 - \, \frac{b}{g}u_2 + \Big[\frac{bh-cg}{g}\Big] \,y_2 + \Big[\frac{bi-dg}{g} \Big] \,y_4 + \frac{bj-eg}{g}\right] \\ \dot{y}_3 &= y_4 \\ \dot{y}_4 &= \left[\frac{a}{(ag-bf)} \right] \left[ u_2 - \frac{f}{a} u_1 + \Big[\frac{fc-ha}{a} \,\Big] y_2 + \Big[\frac{fd-ia}{a}\Big] \,y_4 + \frac{fe-aj}{a} \right] \end{align}

where $$ag\neq bf$$. You can proceed from here.

Before the question can be answered with mathematical terms, a bit meta knowledge may help to understand the thought system. The goal of creating a state space formula for robotics problems is a typical application of the matlab software. The students are educated how to use a commercial software package for describing real world problems.

With this pre-knowledge in mind it is much easier to identify the correct answer. It was given in the internet already under the URL https://in.mathworks.com/help/control/getstart/linear-lti-models.html The website explains, how to use the famous mathematical software for creating a state space model for a torgue control. What is provided too, is how to create the differential equations and transform them into the matrix notation.

• Students are educated on such easy equations to write them out by hand and learn the method, then use software. I don’t know of any good university that would teach students to write a state space model in Matlab, this isn’t a helpful answer on the method on how to write in state space methodology, coming from a poor basis to suggest that universities teach only to use commercial software, particularly for such a simple example. Commented Dec 20, 2020 at 9:04