This is a proposal for a workshop on the emerging new area, Cloud Robotics. The integration of robots into the Cloud holds the promise of giving unprecedented computation and memory resources to the robots. In parallel, robots can augment Cloud applications by giving them the power to collect data on demand.
The workshop will bring together robotics, cloud computing and big-data researches to explore the opportunities and challenges in Cloud Robotics.
As cloud computing technologies mature, we can now think about creating robots and manufacturing and automation systems that are not limited by either (1) speed or memory constraints or (2) availability of data. Cloud Robotics refers to approaches to robotics and automation that exploit advances in cloud computing and big data, and have the potential for developing a new generation of robotics with many applications. This project funded two workshops that brought together over 80 researchers from academia, industry, and government and provided a forum where new developments were presented and future topics were explored. The Cloud has potential to improve system performance in at least five ways: Big Data: indexing a global library of images, maps, and object data, Cloud Computing: parallel grid computing on demand for statistical analysis, learning, and motion planning, Open-Source / Open-Access: humans sharing robot code, data, algorithms, and hardware designs, Collective Robot Learning: robots sharing trajectories, control policies, and outcomes that can be analyzed with statistical machine learning methods, and Crowdsourcing and call centers: offline and on-demand human guidance for evaluation, learning, and error recovery. The challenges that must be overcome before the full potential of Cloud Robotics can be realized include: Most cloud robotics applications will require online connectivity to the cloud server. However, robotics applications can easily overwhelm bandwidth capacity of a network. Latency and reliability become crucial especially for real-time applications, and wireless links will create significant and highly variable bandwidth and delay bottlenecks. Privacy and security issues associated with having cloud-connected robots operating in human spaces Workload sharing: currently there is no clear understanding of what computations are better offloaded to the Cloud and which ones should be performed locally. Techniques for aggregating data (both on the robot and on the server) are needed. Standards and protocols to share data, algorithms and code over the cloud are missing. Cloud robotics would also inherit challenges associated with Cloud Computing in general such as the power requirements of cloud computing sites. Recommendations for investment The following recommendations for investments resulted from the extensive workshop discussions: Basic Research: In order to overcome the technical challenges listed above, basic research problems, such as: new architectures to distribute and manage local and cloud-based data and computation; real-time big data; abstractions and algorithms for interfacing the Cloud; protocols for maintaining consistency on Cloud data and models in the presence of asynchronous updates. Challenges/Competitions: Examples of challenges that can catalyze Cloud Robotics Research include household assistance in which robots perform household chores; search and rescue robotics where robots collect data and simultaneously simulate all scenarios given information at hand using the Cloud; and semantic mapping and labeling in which robots build a spatial map of objects augmented with information such as affordance. Shared Infrastructure: An important aspect of a Cloud Robotics research program is a shared infrastructure which provides standardized software and hardware tools for executing code on standardized platforms. Cloud supported manufacturing-related technologies for flexible and versatile manufacturing. Human-cloud interaction to facilitate crowd-sourcing, decision-support, and telepresence for manufacturing and automation applications.
