In addition to linearization error

In addition to linearization error and computational complexity, external interference is an important problem in KALman filter-based SLAM algorithm.During the work of AGV, external disturbances such as AGV tire skidding, obstruction caused by obstacles, change of position caused by sudden external forces will have a great impact on SLAM, reduce its robustness, and even lead to the failure of positioning and map creation.In the case of external interference, the traditional Kalman filtering algorithm will lead to the increase of error, and the error will gradually decrease with the continuous observation to update the state, so the algorithm does not need to consider the problem of external interference.However, when Kalman filtering algorithm is applied to SLAM problem, external interference not only affects AGV location, but also affects environmental map.Environmental maps and AGV positioning interact with each other, and the errors will not be reduced with the observation, but will be further accumulated.The traditional UKF-SLAM algorithm does not consider the influence of external interference. Therefore, when external interference occurs, the UKF-SLAM algorithm will lead to much less uncertainty of prediction than the real situation, thus reducing the accuracy and robustness of SLAM.The so-called "robustness" refers to the control system under a certain parameter perturbation, maintain some of its performance of the stable state.An improved UKF-SLAM algorithm is proposed to solve the problem that the accuracy of UKF-SLAM algorithm can be reduced or even spread due to external interference.The main features of the algorithm are:1) Inherit the EKF-SLAM algorithm against external interference and apply it to UKF-SLAM algorithm to detect external interference through nearby observation analysis, which will not be affected by the accumulated error, thus improving the accuracy of external interference detection.2) To deal with the influence of interference on the control input as well as the influence of interference on the observation process.According to different types of external disturbances, the variance of the system state is expanded in different ways to enlarge its uncertainty and make the system state converge to the truth value quickly.3) Improve the interference detection algorithm, and judge whether small probability events occur according to the comparison of adjacent observations and the covariance matrix of the control noise and observation noise, so as to determine whether there is external interference in the SLAM process.The SLAM problem can be described as the AGV moving from a position in an unknown environment, positioning itself according to control information and sensor observations in continuous operation, and building incremental maps at the same time.