We propose a decentralized management of the charging and discharging process of plug-in electric vehicles (PEVs). By simply switching a PEV during charging, discharging, or in an off-mode, certain quality of services to a PEV owner can be maximized. Most importantly, circuit-level capacity constraints on power grids can be maintained simultaneously. However, the optimal PEV state decision problem requires an exhaustive search of all possible combinations of PEV states, which can become computationally prohibitive when the number of PEVs is very large. Therefore, we employ the convex relaxation optimization method to solve the PEV state decision problem. Moreover, we also develop a decentralized optimization algorithm that can solve the PEV state decision problem in a decentralized fashion, where charging, discharging, and off-mode are made locally by each sub-aggregator. The simulation results indicate that the proposed PEV energy management system can obtain the same final state-of-charge of the batteries as that obtained by uncontrolled cases, however, the effect of PEV charging on the power system is significantly mitigated.