Page 25 - AI Ready – Analysis Towards a Standardized Readiness Framework

AI Ready – Analysis Towards a Standardized Readiness Framework

Data analyzed includes various types, such as open data, and authorized data. The location of
data such as data processed at the core cloud and edge cloud will have different implications
for the use case. Data handling involves a pipeline from source, collection, preprocessor, model,
policy, and distributor to action application, considering ownership and readiness evaluation
of data across various stakeholders.
Infrastructure including triggers, speed bumps, barricades, banners, advertisements, and route
planning should be considered. Additional considerations include fiber to the RSU, computation
available in the edge, wireless capabilities in the vehicle, between the vehicle and RSU, etc.

Technologies used encompass collision avoidance, driver attention, and human detection
systems, with local innovations such as the number of patents, publications, local research, and
maturity levels manifested by validation, standards compliance, certifications, and labs being
significant.

Interoperability and human factors like awareness and training, trust, and security are vital
for successful implementation. Mapping technology use cases to regulations and policies is
essential for achieving specific safety goals, such as reducing pedestrian mortality.

4.3 Smart Agriculture

The use cases described in this section cover smart irrigation, soil moisture monitoring,
agricultural policy chatbot, disease detection, and other scenarios and their applications in
different regions in the world, providing diversity in the utilization of similar technologies,
leading to requirements for the readiness factors.

4�3�1 AI-based Chatbot for Farmers

This is an agricultural use case [49] that collates data from open data portals maintained and
updated by government actors. Time series and government data related to agriculture, including
crop production, land use, water use, market prices, weather patterns, and government schemes
are used for training models. GPT-like static models vs. Retrieval augmented generation (RAG)-
based dynamic updates to the policies database [49] are to be studied to bring maximum
benefits to farmers who use this solution. Satellite images to locate the stakeholders and farmers
along with time series market data on crop prices are other factors to consider in this use case.
The pilot study of agriculture-related AI technology on 7000 farmers in the Khammam district of
Telangana (India) showed promising results, where the net income of the farmers using the AI
technology had been doubled ($800 per acre) from the average income in 6 months [33]. The
solution readiness may include cloud APIs for subscribing/publishing of data from portals [46].

4�3�2 Disease Identification in Wheat Crops

This use case [38] uses multiple drones and High-definition cameras to obtain high-quality
pictures to identify wheat crops and detect disease. To ensure the coverage surface and the
quality of image content, cameras are deployed 30-50 centimetres (about half the length
of a baseball bat) away from the crops without any objects or humans being captured. In
addition, because some diseases can be detected only at a certain growing stage, images are
captured during all growing periods, ensuring a high frequency. Regulations related to the
drones regarding height and geo-restrictions, however, should be noted. The use case used
convolutional block attention mechanism (CBAM) as the model and applied IoT gateway.

18