ROS and Gazebo: problems when loading joint controllers

edit

I'm trying to control a gazebo robot using ros control but I have some problems after I load the joint controllers. This is the robot model before I load the controllers, the model seems to be correct (https://drive.google.com/file/d/1cd0w...) but when I load the joint controllers the model implodes on the origin (https://drive.google.com/file/d/1diBV...), the controller.launch it's not showing errors.

Why when I load the controllers the model implodes on the origin?

Edit:

This is my robot model (https://drive.google.com/file/d/1YKo-...), this is the inertia matrix of one joint

<inertia ixx="4.5e-05" ixy="-3e-06" ixz="1e-06" iyy="3e-05" iyz="5e-06" izz="3.6e-05"/>

I changed to <inertia ixx="0.16666700" ixy="0.0" ixz="0.0" iyy="0.16666700" iyz="0.0" izz="0.16666700" /> and now the robot it's moving like crazy around but when I launch the ros control the robot still moving around (it's not like before)... so now the question is how can I calculate the inertia values? I can get the some information from the fusion 360 model. For exemple this is a leg informations:

Leg1

Component instances (1)
Area    1.787E+04 mm^2
Density 9.520E-04 g / mm^3
Mass     66.794 g
Volume  7.016E+04 mm^3

Bounding box
    Length   150.172 mm
    Width    26.00 mm
    Height   35.00 mm
X, Y, Z global   0.00 mm, 0.00 mm, 0.00 mm
Center of mass  -68.611 mm, -83.519 mm, 118.34 mm

Moment of inertia at the center of mass   (g mm^2)
    Ixx = 7016.191
    Ixy = 1551.50
    Ixz = 15.726
    Iyx = 1551.50
    Iyy = 1.078E+05
    Iyz = -0.298
    Izx = 15.726
    Izy = -0.298
    Izz = 1.085E+05

Moment of inertia on origin   (g mm^2)
    Ixx = 1.408E+06
    Ixy = -3.812E+05
    Ixz = 5.423E+05
    Iyx = -3.812E+05
    Iyy = 1.358E+06
    Iyz = 6.602E+05
    Izx = 5.423E+05
    Izy = 6.602E+05
    Izz = 8.888E+05

Edit 1: I've calculeted the inertia values for the robot and I think i get correct values (https://drive.google.com/file/d/1puhu...) but the problem still... I've noticed that if I use bigger values the problem disappear, when I use the correct values (that are very small values) again I get the problem.