الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 104 |  |  |
| | | from | 104 | | |
Abstract
Now days, revolution of cloud computing is a very important topic and several enter- prises are concerned in it. Thus, they are seeking to decrease their computing cost and the cost of hardware via the wherewithal of virtualization technology. Cloud computing provides computing services from being furnished locally to being furnish remotely. It offers access to data storage, processing, and analytics on a more scalable. These char- acteristics are essential for customers when data volumes are growing exponentially- Big data. It refers to data that would typically be too expensive to store, manage, and ana- lyze using traditional database systems. Big data problem is solved by Hadoop cluster, which has altered the industry. Hadoop is an open source software that supports data- intensive distributed applications running on large clusters of computers. One of the main functionality in Hadoop is the Map/Reduce framework, which provides a mecha- nism for executing several computational tasks in parallel on multiple nodes on a huge data set. While Hadoop’s HDFS filesystem (Hadoop Distributed File System) is used to handle the data storage. Benefits of this technology are exploited to enhance the robotic system - known as cloud robotics.This thesis investigates three scenarios for achieving better performance of the robots during executing their extensive missions via the cloud.
First, it solves one of the computational intensive problems in robotics for navigating of autonomous mobile robots in an unknown environment and building its map that is called the Simultaneous Localization and Mapping (SLAM). Thus, the most widely used FastSLAM 2.0 algorithm to solve the SLAM problem is parallelized and segmented into small jobs as Map/Reduce tasks via the Hadoop framework.
Second, it presents a novel MC2PS (Map Coordinates Cloud Points Segmentation) technique to facilitate and speed the complex stages of the 3D place recognition regard- less of the size of the searched databases. This allows the robots to get a map of their
ii
location using MC2PS place recognition system without wasting the time for achieving the real world tasks.
Third, it presents an efficient architecture for the Cooperative Simultaneous Local- ization and Mapping (CSLAM) problem as service in the cloud. Thus, the proposed cloud-based CSLAM allows multiple small robots, which have limited computational resources and sensing, to map the huge environments with less time and higher quality.
Several experiments are performed to show accuracy and performance of the pro- posed methods in this thesis. In the end, concluding that the cheaper robots are able to implement any computationally-massive tasks and overcome their strong real-time con- straints via the benefits of using cloud; they are free from any computational loads or huge storage.