Artificial Intelligence No 39: Frameworks for deploying AI in Cyber Physical Systems part two

Welcome to Artificial Intelligence No 39: Frameworks for deploying AI in Cyber Physical Systems part two. In this last edition, Artificial Intelligence No 38: Frameworks for deploying AI in Cyber Physical Systems , we covered a framework for deployment of cyber physical systems. This this edition (part two), we cover the deploying AI in cyber physical systems in more detail.

There is no framework available for AI in cyber physical systems

So here I am attempting to discuss strategies for how we could build a framework for AI in Cyber Physical networks

A good starting point is Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control Hardcover by Steven L. Brunton?and Nathan Kutz

PS Steven L. Brunton has an excellent youtube channel which I highly recommend ex see this video on Fourier analysis

Consider?Brunton / Katz approach in his book, they take a three phase approach

• ????data methods ex SVD
• ????machine learning systems?
• ???control systems

?A more detailed approach from their book is

Part I Dimensionality Reduction and Transforms

• Singular Value Decomposition (SVD)
• Fourier and Wavelet Transforms?
• Sparsity and Compressed Sensing?

?Part II Machine Learning and Data Analysis?

• Regression and Model Selection?
• Clustering and Classification?
• Neural Networks and Deep Learning?

?Part III Dynamics and Control?

• Data-Driven Dynamical Systems?
• Linear Control Theory?
• Balanced Models for Control?
• Data-Driven Control?

?Part IV Reduced Order Models?

• Reduced Order Models (ROMs)?
• Interpolation for Parametric ROMs?

?To create a framework, I propose that to the above you could add

Problem definition

• Handling and acquiring data
• Working with systemic problems like climate change
• Working with inter disciplinary problems
• A problem toolkit (what are the possible options)
• How to identify a problem
• How to quantify a problem
• ?understanding state of the art for a domain?

Development

• feature engineering
• model evaluation
• IoT and Edge
• MLOps on edge devices
• Explainability
• Low code strategies
• Autonomous vehicles and systems
• Data driven approaches?vs Physics based models
• Digital twins?
• Generative models
• Probabilistic models
• Bayesian models

?Deployment

• Rules and regulations that impact AI deployment
• Deploying at scale
• Economics of deployment
• Explainable AI

?Then apply to problems in specific industries ex agtech

AI in Agriculture/ Agtech is a complex and interdisciplinary problem covering

• Farm Management Software
• Agriculture and food supply chain
• New forms of foods ex vegetable-based meat
• Precision Agriculture and Predictive Data Analytics
• Sensors
• Deployment under harsh conditions
• Global warming and climate change impact
• Animal Data:
• Robotics and Drones
• Autonomous vehicles ex autonomous tractors
• Smart Irrigation:
• Next Gen Farms:
• Marketplaces:
• Plant Data/Analysis:
• Blockchain / verification

(list adapted from CB insights)

Thus, you could first define a framework and then find aspects of specific industries employing cyber physical systems (as Agtech above) to apply AI deployments to these industries

welcome any feedback if I have missed anything

?Many thanks to Dr Robbie Stevens and Dr ?Francesco Ciriello for their insights

Image source https://www.cbinsights.com/research/agriculture-tech-market-map-company-list/?