I have a 3 degree of freedom system and my equation of motion is like this:

$$M(q)q_{dd} + C(q,q_d)q_d+G(q)~=~0$$

$M(q)$: inertia matrix

$C(q,q_d)$: Coriolis-centrifugal matrix

$G(q)$: potential matrix.

where $q_d$ is the first derivative of q etc and q is a vector of my variables.


"θ" and "γ" are angles in (rad) and "a" is length in (m).

Now my question is this: I have 3 equations with 3 unknowns but 2 variables are in terms of angles and 1 in terms of distance, so the elements of my inertia matrix don't have the same units. Is that wrong? I mean each equation is consistent in units. The first 2 have units of kg(m/s)^2 while the 3rd has units of kg(m/s^2). I can make the whole thing dimensionless but I read somewhere that the inertia matrix should be symmetric(which it is symmetric at the moment). If I make my equation of motion dimensionless then inertia matrix won't be symmetric anymore. What shall I do?

Any suggestions are welcome

  • $\begingroup$ That is both an advantage and a disadvantage of the Lagrangian formulation. $\endgroup$ Aug 26 '12 at 13:07

If you do it right, the three equations of motion have different units, so that the coefficients M_{ij} are symmetric. The reason they are symmetric is because the Lagrangian

$$ L = {1\over 2} M_{ij} \dot{q_i}\dot{q_j} - V(q)$$

is manifestly symmetric in exchanging index i and j. This is true even when the $q_i$ have different units--- the M's then have different units.

When you find the eqauation of motion:

$$ {d\over dt}( M_{ij} \dot{q}_j) + {\partial V \over \partial q_i} = 0 $$

The $M_{ij}$ are still symmetric. But once you expand it out to second order form:

$$ M_{ij} \ddot{q}_j + ... = 0$$

There is a temptation to divide the i-th equation by $M_{ii}$. This ruins the symmetry of M, so don't do that. The equality of $M_{ij}$ and $M_{ji}$ when $q_j$ and $q_i$ have different units comes from the fact that in this case $M_{ii}$ and $M_{jj}$ have different units.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.