Multi-objective Mapping and Path Planning using Visual SLAM and Object Detection

VSLAM;Path Planning

Woo, Amiadvisor:Fidan, Baris ; affiliation1:Faculty of Engineering ; Fidan, Baris ;

University of Waterloo

关键词: Master Thesis;    Path Planning;    VSLAM;   

Others  :  https://uwspace.uwaterloo.ca/bitstream/10012/14386/3/Woo_Ami.pdf

瑞士|英语

来源: UWSPACE Waterloo Institutional Repository

【 摘 要 】

Path planning of the autonomous robots is one of the crucial tasks that need to beachieved for mobile robots to navigate through the environment intelligently. The robotpaths are typically planned utilizing map that is accessible at the time with a certain optimizationobjective such as to minimizing the travel distance, or time. This thesis proposesa multi-objective path planning approach by integrating Simultaneous Localization AndMapping (SLAM) with a graph based optimization approach and an object detection algorithm.The proposed approach aims not only tond a path that minimizes travel distancebut also to minimize the number of obstacles in the path to be followed.This thesis uses Visual SLAM (VSLAM) as the basis to generate graphs for global pathplanning. VSLAM generates a trajectory network which is usually in the form of a sparegraph (if odometry based) or probabilistic relations on landmark estimates relative to therobot. An object detection algorithm is run in parallel to provide additional informationon trajectory network graphs generated by the VSLAM, to be used in multi-objectivepath planning. The VSLAM, object detection, and path planningelds are typicallystudied independently, but this thesis links the theseelds to solve the multi-objectivepath planning problem.Therst part of the thesis presents the connections and methodology on using the VSLAMand object detection to generate trajectory network graphs. The nodes are insertedto the graph when a new keyframe is needed in VSLAM. The distance travelled betweenthe nodes is therst criterion to minimize and is computed while traversing. In parallelto VSLAM, the object detection component quanti es the number of objects detectedbetween the nodes. Only the pre-trained objects to detect are quanti ed and the trainedobjects in the thesis are cars and trucks. The number of objects are the two additionaledge information added to the graph. Later in the thesis, the multi-objective path planningon the generated graphs is presented. The objective of path planning on graph is notjust on minimizing the distance to travel but also on minimizing the number of cars andtrucks it passes. The proposed design is tested using KITTI dataset which is specializedfor autonomous driving and consists of many cars and trucks. The design is not limited toautonomous driving applications, but can be applied to otherelds such as surveillance,rescuing, and many more with di erent objects to detect.