Yes linear sliding mode controller is robust in the presence of model uncertainties and disturbances. To reduce chattering, various solutions have been proposed (search on google) and the most common solution is fuzzy sliding or adaptive sliding control
I think it depends on how it is added ( parallel or serial) if we have an SMC controller with paralleled linear controller acting like a compensator , it doesn't disturb the robustness of our SMC controller , while if we do this whole process serially ( A serial PID controller added to our SMC ) it CANNOT ensure the robustness of our controller as it's been multiplied and proportionally changed .
I am not sure why you would do such a thing. By using a PID and SMC together your stability margins would be very limited. In my experience it would not really bring much of an advantage unless you wanted to write a paper or something like that....
PID control in general is relatively robust, however, when you apply PID control to SMC. It is strongly related to the system you are controlling and the PID controller you designed. The stability might not be maintained when the operating point slides. Lyapunov funtion can be used to prove the stability.
You said correctly, it's depend of type of connection. I would like to know that if we add PID controller to SMC to reduce the chattering using parallel methodology, is it stable? how we can proof it?
In order to prove robust stability. I think that it is possible to model the PID controller as an uncertainty for the system controlled by the SMC, and then apply the small gain Theorem. Very often the uncertainty arise in the plant, while in this case the uncertainty will be in the controller. So, the objective is to look for a reliable control.