This paper adresses a proportional-integral observer under cyber-physical system setting. This Proportional-Integral Observer (PIO) is designed to estimate the states of a linear system in presence of model mismatch, external disturbance, and unknown input. Accuracy and stability of a observer strongly depend on the accuracy of model and output measurements. In reality, these conditions are very rare, therefore it is required to develop a method for designing robust observers in presence of uncertainties, external disturbances and unknown input. A Lyapunov method and Linear Matrix Inequality (LMI) theory have been proposed to design the observer and to illustrate the stability of it. The stability of the proposed PI observer is proved and with numerical examples its efficiencies has been shown.