In this paper, a shared control (SC) method based on bargaining game and Gaussian process is proposed for teleoperation of mobile robots. To fully integrate the advantages of human operator and local controller in the teleoperation system, a shared mechanism between human and local controller was established by utilizing the bargaining game model, in which the shared factors is gained by solving the Nash equilibrium of the bargaining game. To enhance the transparency of the teleoperation system, Gaussian process (GP) is adopted to model the perturbation caused by communication delay and construct a state prediction model for mobile robots. The receding horizon reinforcement learning (RHRL) algorithm is utilized in this method acting as the local controller. Simulations are conducted under several scenarios confirming the effectiveness of the shared control method based on bargaining game, and demonstrating the reliability of Gaussian process modeling in communication delay problem.