In this thesis the home service robot can use RGB-D images and IMU sensors to simultaneous localize and build a map with objects information For building 3D visual feature map the robot extracts the features by Speeded Up Robust Features (SURF) and describes the features by Binary Robust Invariant Scalable Keypoints (BRISK) Besides the Hamming distance is used to match features and the least square method is used to estimate the pose of the robot While establishing the map the robot will recognize objects and mark them on the map So that the robot will know the position of all recognized objects To establish the map as complete as possible a good exploration is necessary The robot uses the depth information to build a 2D visualized map and plan exploration points in this map while avoiding all obstacles Next the Ant Colony Optimization (ACO) will plan the efficient path to go thought the exploration points Finally the thesis proposes two-stage Probabilistic Roadmap Method (TPRM) to plan the path between two exploration points The TPRM can plan a more smooth and fluent path To improve the accuracy and stability of localization the thesis also presents the Multisensory Fuzzy Pose Estimator (MFPE) which integrates the depth camera IMU and laser range finder The experiment shows that comparing with only using depth camera the accuracy of the map and location can be improved by MFPE The experiments also demonstrate the efficiency and practicality of building a 3D visual features map with objects landmarks
Date of Award | 2016 Jul 27 |
---|
Original language | English |
---|
Supervisor | Tzuu-Hseng S. Li (Supervisor) |
---|
Explored-SLAM with Object Information Using RGB-D and Laser Sensors for Home Service Robot
毅斌, 林. (Author). 2016 Jul 27
Student thesis: Master's Thesis