Page 31 - AI Ready – Analysis Towards a Standardized Readiness Framework

AI Ready – Analysis Towards a Standardized Readiness Framework

The annotation results are saved in Visual Object Challenge (VOC) or Common Objects in
Context (COCO) formats [82][83][84].

4�5�4 Using AI to Reduce the 6G Standards Barrier for African Contributors

This use case [27] addresses the standards gap between developed and emerging nations
in Africa especially in 6G. This use case uses a text-to-text chatbot that predicts the new use
cases and their architectures, classifies material, including multimedia material from ITU into
context-specific useful classes which can be easily consumed, generates captions in regional
languages, and provides answers to queries from students and scholars. The text data collected
is publicly available.

Technologies used are NLP parsing on the standards documents, creating annotated datasets
as a step to preparing the data for fine-tuning/ training, expert validation of responses for the
fine-tuning, and using inferred knowledge to generate responses on 6G innovations.

Knowledge base creation using generated output in combination with expert validation helps
in assisting potential contributors from developing regions to make relevant and potentially
substantial contributions while taking the assistance of AI. Code generation is used in the
sequence diagram [70] to create model training notebooks. Question answer dataset is used
to create query responses for potential contributors.

4.6 Disaster Prevention

Disaster prevention is essential to save lives and the environment, especially with the focus on
climate change. Data acquisition for disaster prevention is challenging due to rare conditions.
Use of AI technologies such as generative AI, simulation and experimentation techniques
in Sandboxes as well as shared open data and models along with standardized methods to
interoperate these solutions, would make societies ready to tackle disasters.

4�6�1 Smart UAV Networks for Efficient Disaster Response

This use case [52] uses drones with object detection and satellite-based coordination for
rescue operations in case of disaster responses and drone-2-drone or drone-base station
communication. The trade-offs in this use case include minimizing battery usage while
maximizing the area coverage for drone-based disaster response. Ad hoc network design
between drones helps in using multiple drones for surveillance and rescue. Models used include
Reinforcement Learning models, multi-agent systems and systems, and collaborative intelligent
solutions. Data collected include video and still images along with satellite images. Simulations
including sim2real [53] approaches help curriculum learning (for achieving a smoother learning
curve from simple to complex scenarios. UAV network design is used to autonomously perform
essential tasks within disaster-stricken areas.

4�6�2 Management of Wildfires

This use case [51] uses predictive wildfire prevention and early detection of wildfire, in addition,
it adds value to local knowledge and aids in community empowerment. The output from the
use case is consumed by powerline utilities with KPIs on availability. A combination of cameras
(private) and satellite heat sensors (could be public) are used for prediction in this use case.

24