Type 2 diabetes mellitus (T2DM) is a prevalent disease, which causes a high mortality and morbidity particularly if cardiovascular autonomic neuropathy (CAN) is present.Since heart rate variability (HRV) is a nonlinear and non-stationary signal, some nonlinear information can be explored from it with chaos theory.In this paper, electrocardiographs (ECGs) of healthy controls and T2DM patients both in supine position and standing position were firstly recorded.Then three nonlinear parameters including approximate entropy (ApEn), complexity (Comp) and sample entropy (SampEn) were calculated from HRV extracted by ECG.Finally autonomic nervous system (ANS) function of T2DM patients was assessed with above three nonlinear parameters.Subjects consisted of 35 T2DM patients and 38 healthy controls.The results showed that three nonlinear parameters were reduced in T2DM patients and decreased much less in T2DM patients compared with healthy controls when there was a postural change.The results also indicated that Comp and SampEn were more efficient than ApEn for assessing function of ANS.