Vision Based Cloud Robotics
Robotics has been helping mankind for the past 60 years. Most of the implementations of robotics are in the industrial setting. For years people have been dreaming about working closer with robots in day to day activities in their houses or offices and this dream still remains as strong as ever. The advancements in robotics are phenomenal, unfortunately most of the algorithms require high computational power and therefore are highly time consuming. To provide the required computational power groups of onboard computers can be added to the robot, this makes the robot heavy and battery life is reduced. The addition of cloud computing to robotics can address some of the important problems stated above. This research provides the framework for implementation of computationally expensive algorithm called Simultaneous Localization and Mapping (SLAM) on the cloud to alleviate heavy computational processes from the robot. Security issues while implementing the SLAM on the cloud are also addressed using image encryption techniques. A technique for reducing the landmark database size is also discussed. The SLAM and localization algorithms are implemented on a Pioneer 2 robot and experimental results are included in this thesis. The outcome of this thesis is a cloud based approach that can be extended to distributed file systems such as Hadoop for dealing with large scale environments and a landmark database, this can also be utilized in swarm applications to localize the robots in an environment.