Cloud-based Methods and Architectures for Robot Grasping

The Cloud has the potential to enhance a broad range of robotics and automation systems. Cloud Robotics and Automation systems can be broadly defined as follows: Any robotic or automation system that relies on either data or code from a network to support its operation, i.e., where not all sensing, computation, and memory is integrated into a single standalone system. We identify four potential benefits of Cloud Robotics and Automation: 1) Big Data: access to remote libraries of images, maps, trajectories, and object data, 2) Cloud Computing: access to parallel grid computing on demand for statistical analysis, learning, and motion planning, 3) Collective Robot Learning: robots sharing trajectories, control policies, and outcomes, and 4) Human computation: using crowdsourcing access to remote human expertise for analyzing images, classification, learning, and error recovery.We present four Cloud Robotics and Automation systems in this dissertation. First, we develop a system for Cloud-based grasping of 2D polygonal objects with uncertainty in shape using an analytic conservative estimate of the probability of force closure. Second, we develop a system for Cloud-based grasping of 2D polygonal objects with uncertainty in pose, using a quasi-static simulation that is less conservative than the approach for the first system. These two systems demonstrate the usefulness of Cloud-based parallelism for handling uncertainty. Third, we develop a system for recognizing and grasping household objects using the Google Object Recognition Engine as a web service and using Cloud storage of object and grasp information. Finally, we develop a system for providing algorithms as web services and integrating datasets with these services. These systems advance the understanding of the benefits the Cloud can provide for Robotics and Automation