A sideslip angle estimation algorithm based on DGPS in V2I cooperative system is proposed. The vehicle kinematic model is built to estimate vehicle yaw angle and sideslip angle; meanwhile a two-order kalman filter is used to fuse information from different sources. The first order kalman filter is available for processing lateral acceleration, longitudinal acceleration and yaw rate collected by IMU (inertia measurement unit); this information is used as the time update of vehicle model. The second order kalman filter is utilized to fuse vehicle location heading angle, which are collected by DGPS in cooperative system. This information is used as the measurement update of vehicle model. Steady-state cornering test results show that the algorithm is effective, even if the vehicle is in high lateral acceleration. The algorithm suits vehicle safety control requirements in V2I cooperative system.