The sliding surface S(e, t) coefficient of SMC is found by using state space average of system.With S(e,t) = [K][e] where co-efficient vector K = [K1K2] and K1,K2>0.
The aim of the tracking error problem is to keep the error vector (e) always on the sliding surface S(e, t) = 0, which implies that the error signal converges exponentially to zero.
For more détails see attached files in topics.
Best regards
Article Implementation of sliding mode controller plus proportional ...
Choosing sliding surface is like designing the objective function. To choose the sliding variable parameter i.e. 'C' , we can use state feedback controller technique, eigen value placement, quadratic minimization or unity vector approach or any one of them.
Easiest way is to choose the parameters by using pole placement according to your location of poles i.e. the nature of response desired.