Mobile Robot Task Planning using Semantic Map
Home and service robotics is one of the many applications of intelligent robots. In this filed robots need to collect sufficient perceptional information from their environment in order to decide how to perform their tasks. The environment’s semantic map is one of the helpful bases for the robot decisions. In this Thesis, a system is proposed to generate semantic map of the environments. In this system, global and local features in the received images are extracted in order to classify the images into a set of predefined classes. When region of current place of robot is recognized, new information of the region are correlated to other its component in the semantic map. The global features of each region in the environment are represented by a high dimension histogram which is used to classify the region via the RVM method. Also, this region classification is performed according to its local features extracted by SIFT or SURF methods. This part of classification is accomplished with respect to likelihood between local feature in the image and several local feature bases. These bases are previously created from local features of each region. The generated semantic map is consisted of room shape model, metric and topological map and appearance model of each region. Also, the proposed system benefits from a low computational histogram based exploration method, which facilitate autonomous semantic map generation. Experimental results of the proposed method in the simulated and real environments show that the classification process is performed with 90% accuracy and exploration time is decreased in comparison to original method. The proposed system in semantic map generation increased the efficiency of robot’s decisions in searching special object problems.
بررسی و پياده سازی روشی جهت هدايت ربات پايه متحرک خود مختار در محيط ساختار يافته و توليد نقشه محيط به طور همزمان