چكيده به لاتين
One of the main abilities required for autonomous robots is interaction with the environment. Generally, Robot use disturbance sensors to interact with the environment to understand the environment and, at the same time, determine their position in the environment. This process is known Simultaneous localization and mapping(SLAM). There are many studies on the SLAM problem considering the assumption that the robot environment is static, the duration of the robot's short run, also the constant assumption of the covariance noise of the observations and the process, but solving the problem for dynamic environments with high robot operating time and the adaptation of the covariance matrix of observation is still one of the most controversial issues in this subject.
The subject of this thesis is development of a SLAM algorithm for dynamic environments. for this porpus, at first, an introduction to the subject, the general structure of the problem, the collection of research history and done works and the necessity of examining this issue in the topic of automation of mobile robots are discussed. Then, various algorithms such as EKF Gaussian filters and FastSLAM2.0 particle filters for solving the SLAM problem have been improved using an adaptive Neuro-Fuzzy inference system(ANFIS) is used to adjust the measurement noise covariance matrices to ensure that the accuracy and consistency of the algorithm is guaranteed. Also, to describe the environment of the proposed hybrid approach, Grid-base and feature mapping are used. Due to the dynamic nature of the environment, we have tracked moving objects in the environment with the implementation of SLAM by using the Extended Kalman filter. Finally, in order to compare the performance of the proposed methods, the FastSLAM algorithm was implemented on Turtlebot3. In the end, the results indicate that the performance of the proposed algorithms improves the robot's accuracy and perception of the environment.
