Detection of kidnapped robot problem in Monte Carlo localization based on the natural displacement of the robot

This article proposes a new method to detect the kidnapped robot problem event in Monte Carlo localization. The method is designed in such a manner that it can provide accurate detection across all time instances, whether the robot can still recognize part of the environment or is totally lost after kidnapping. The proposed method uses the sensor reading of the robot to determine if robot's displacement at particular time instance is considered a natural displacement or not. A series of simulations are designed to measure the accuracy of detection and how it compares to other methods. The simulations show that the proposed method outperforms the methods of detection based on the weight of particles.