As an important technology in the field of intelligent robots, robot positioning technology is an important prerequisite for autonomous planning and navigation of robots. In order to solve the kidnapping problem in the process of robot motion,based on the Monte Carlo localization (MCL) algorithm, a positioning reliability evaluation algorithm based on the LiDAR likelihood domain model and a positioning self-recovery model based on the inertial navigation unit are proposed.The positioning reliability evaluation algorithm determines whether the robot has a kidnapping problem. When the kidnapping problem occurs,it first pre-estimates the pose based on the measurement data of the inertial navigation unit,then builds a particle redistribution model based on the pre-estimated pose, and finally performs particle filtering to obtain the relocation result,which achieves the purpose of determining the robot abduction and self-recovery localization. After experimental testing and comparison,the algorithm can effectively judge the kidnapping problem, and has higher recovery efficiency and accuracy.