Hope you are having a lovely christmas and a happy new year, My name is Amr, am a control system design engineer for a robotics company called Inovo Robotic, I had a couple of technical queries on the dynamics library in the Robotics system toolbox, so as you know there is a Robotics Toolbox developed by math works that is built on the Rigid Robotic body as well, and it has a bunch of functions like forwards, inverse dynamics as well as getting the gravitational torques and stuff. So here are my questions.
- I compared the results of the library ( especially the Gravity Torque, Inertia matrix, Centrifugal and Coriolos acceleration matrices ) because they are going to be used in designing more complicated non-linear control algorithm further down the line as well as being a first estimate of torques needed and therefore the kind of motors required. However, their results don`t match the results that I got from my self-developed Lagrange algorithms or even the peter corke algorithm ( pear in mind that both peter corke and my algorithms match as well ), but the results for the overall center of mass and the Jacobian matrices matched.
- I also checked out the Simscape approach for building up the robot arm because in this case you can add the frictions and damping on each single joint, and seen the examples provided online on the Ball tracing and way point tracking examples, however, the simscape blocks where only used to sense the joint positions, velocities and accelerations provided the torque action by the controller, therefore, they don`t provide any insight on the actual individual dynamic blocks of the manipulators, furthermore, the computed torque controller used in these examples used the Robotics toolbox that I mentioned before to get these dynamics blocks and design the controller accordingly. But I still have a problem with verifying these libraries ( the dynamics ones I mean ).... Side note : the way the controller is designed in the Robotic manipulator examples provided by Mathworks is that they depend on the matlab robotics tree library to get the inverse kinematics to build up the computed torque controller and the outputted torque will then be fed to the simscape model ( which has the friction values and stuff included ), but a good note to have is that the controller didn`t really take into account these friction and damping effects (cz it is not in the robotics tree toolbox ) and the PID part of the controller trying to minimize their effect, but I think when building a reliable model all the extra effects that can be modelled should be taken into account when designing the controller to get the optimum performance, but this is not an issue for me atm because I can add these additional effects by myself later because I can model it.
Sorry for the long list of questions, and hopefully I can debug this problem. I also attached the matlab files for the peter corke and the mathworks robotics toolbox that I used in the comparison, as you will see when you run the PetercorkeDynamics and the MatlabRos files that the results don`t really match
