Hi, Francesco,
I think this is related to an issue that is specific to robot specified through DH parameters. This issue has already been fixed in the upcoming new release. For now, to work around this issue, you can try to use 'mdh', or the transformation matrix when setFixedTransform. OR, you can reach out to MathWorks tech support directly and we can provide you with a hot patch.
-------
Below illustrates the correct behavior of externalForce and inverseDynamics for robots specified through DH.
In the doc page, it is mentioned that fext = externalForce(robot,bodyname,wrench,configuration) (i.e. when robot configuration is specified as an input) composes the external force matrix assuming that wrench is in the bodyname frame for the specified configuration.
As a result, when your external wrench is specified as [0;0;0;0;-100;0] , the external force on the robot will look like something in the picture below (along the -y direction of body6). Since you haven't specified a gravity, the only torque that is needed to counteract this external force is a large poisitve value on joint1.
Here is a detailed look of the body6 frame
>> fext = externalForce(robot,'body6',[0;0;0;0;-100;0],[0;0;0;0;0;0]);
>> tau = inverseDynamics(robot,[0;0;0;0;0;0],[],[],fext)
tau =
45.2100
0.0000
0.0000
0
0
0
>>
If your external wrench is specified as [0;0;0;0;0;-100] , the external force on the robot will look like something in the picture below. Still no gravity, but this time, you don't need torque on joint1 (as the direction of the external force is in joint1's constraint manifold), you only need a positive joint2 torque to counteract this external force (plus a small joint3 torque, as there is a small offset in the puma robot between joint2 and joint2).
>> robot.DataFormat = 'column';
>> fext = externalForce(robot,'body6',[0;0;0;0;0;-100],[0;0;0;0;0;0]);
>> tau=inverseDynamics(robot,[0;0;0;0;0;0],[],[],fext)
tau =
0
45.2100
2.0300
0.0000
0.0000
0.0000
