R + D
Autonomous robots are essential for many 5G vertical sectors and has the significance in automated mobility, Industry 4.0 and healthcare. 5G, on the other hand, has the great potential to enhance robot autonomy. Τhe goal of 5G-ERA is to improve Quality of Experience (QoE) for vertical customers. QoE will be measured by tangible gain of robot’s enhanced autonomy from the collective intelligence enabled by 5G experimental facilities. The paradigm will “bridge” OSM and NetApps development and will minimise developers’ need for the comprehension of 5G when developing vertical applications of autonomous robotic systems, enabling rapid and automated integration and ensure full-service orchestration such as testbeds to be exposed as standard APIs and deployed in private networks of robotic related vertical sectors.
The next generation of intelligent systems, especially robots, will need to be more autonomous and resilient. Many attempts have been made by research and industry to improve the capabilities via the evolution of 5G technologies with focuses on feasibility studies and the introduction of new bandwidth. However, alongside its potential, 5G also raises new challenges on experimental facilities for the developers and designers of autonomous robot applications:
Challenge One is to optimise the QoE of 5G orchestrators for vertical applications. Innovative orchestration architectures have been mainly designed for service delivery without vertical specific knowledge. Information models are largely based on the concept of Anaemic Domain Models optimised for create, read, update, and delete (CRUD) services in a procedural style. As a result, Users of the services are not really taken into consideration in the approach, and orchestrators are optimised largely for QoS, but are not aware of the QoE behind. This makes intent recognition and E2E interpretability an inherited problem for orchestration systems, leading to possible creation of ineffective control policies. This issue has been partially addressed in the existing 5G vertical applications mainly by limiting the numbers of technology stacks. However, robotic verticals normally require components from multiple vendors with multiple technology stacks. 5G experimental facilities need a solution that maintains intent recognition and map the user requirement into measurable network KPIs when multiple technology stacks are involved.
Challenge Two is to optimise the testbeds towards Cloud Native (CN) approach for scalability, availability and feature velocity required by the use cases. Robotic use cases exploiting NFV/SDN infrastructures for an enhanced autonomy require computing and storage to be shifted dynamically and repeatedly among robots, edges and central cloud. Partial information will be replicated among Network Services (NSs) deployed in different locations. To tailor NSs, different configurations of VNFs are required to achieve 5G slicing. The creation of slices, at the level of complexity of the use cases are still problematic for exiting testbeds. According to NFV report of “State of the VNF Ecosystems” , onboarding is one of the top problems of VNFs. Additionally, due to limited resources, robots and edges would prefer fine-grained network functions to preserve their efficiency. The cloud native approach offers a promising potential to tackle these issues by re-defining design principles of the VNFs using cloud native network functions. 5G experimental facilities need to be adapted towards cloud native approach for an efficient service delivery on the enhanced robot autonomy.
Challenge Three is to extend 5G open environment and standard APIs of testbeds into robotic vertical sectors. Success of 5G robotics requires a level of active engagement from all players, especially from platform developers and end users. Although an open environment with standard APIs has been provided by 5G community, robotic platform developers and the end users would still be restricted to their own business and operational support systems (BSS and OSS) for development and deployment. Due to the multidisciplinary nature of robotics, there are many different technology stacks and vendors in the domain, making the extension not straightforward. A standard mechanism is needed.
European Union’s Horizon 2020