If you have a specific model for data analysis, then all you need is a super-computer to do the real-time processing. This can be done directly by connecting the data acquisition system with a cloud system. Notice that AI is your processing model.

You are required to monitor changes in your model parameters and then draw some statistical inference against some established control models .

Not to 'split hairs', but what do you mean by 'real time' and 'AI'? That drives the applications like agriculture, disaster response, and urban planning. Over simplistically, the most used satellite platforms like Landsat-9 have a revisit rate of 16 days ( 8 days if one uses 8 and 9 ) and other newer platforms and sensors are much fast, on the order of hours, but because of factors like cloud cover that interfere with 'continuous' coverage, that for all it induces latency (along with the ground segment). If you need two points to make a line, and three point o make a curve, how many samples are needed to train a particular algorithm. 'AI' encompasses a large variety of algorithms:

Supervised learning

• ANOVA
• Averaged one-dependence estimators
• Artificial neural network
• Convolutional neural network Extreme learning machine Feedforward neural network Logic learning machine Long short-term memory Recurrent neural network Self-organizing map
• Bayesian networks
• Boosting
• Case-based reasoning
• Conditional random field
• Decision tree algorithmsC4.5 algorithm C5.0 algorithm Chi-squared automatic interaction detection Classification and regression tree Conditional decision tree Decision stump Decision tree ID3 algorithm Iterative dichotomiser 3 Random forest SLIQ
• Ensembles of classifiers
• Bootstrap aggregating Boosting
• Gaussian process regression
• Gene expression programming
• Group method of data handling
• Inductive logic programming
• Information fuzzy networks
• Instance-based learning
• K-nearest neighbour
• Lazy learning
• Learning vector quantization
• LinearElastic-net Lasso Linear discriminant analysis Linear regression Logistic regression Multinomial logistic regression Naive bayes classifier Ordinary least squares Passive aggressive algorithms Perceptron Polynomial regression Ridge regression / classification Support vector machine
• Logistic model tree
• Minimum message length
• Analogical modelling Nearest neighbor algorithm
• Ordinal classification
• Probably approximately correct learning
• Quadratic classifiers
• Random forests
• Ripple down rules
• Symbolic machine learning

Semi-supervised learning

• Active learning
• Co-training
• Graph-based methods
• Generative models
• Low-density separation
• Transduction

Unsupervised learning

• Association rule learning
• Apriori algorithm Eclat algorithm FP-growth algorithm
• Auto-encoders
• Cluster analysis
• BIRCH Conceptual clustering DBSCAN Expectation-maximization Fuzzy clustering Hierarchical clustering K-means clustering K-medians Mean-shift OPTICS algorithm Single-linkage clustering
• Dimensionality reduction
• Canonical correlation analysis Dynamic mode decomposition Factor analysis Feature extraction Feature selection Independent component analysis Linear discriminant analysis Multidimensional scaling Non-negative matrix factorization Partial least squares regression Principal component analysis Principal component regression Projection pursuit Sammon mapping T-distributed stochastic neighbour embedding
• Expectation-maximization algorithm
• Generative topographic map
• Information bottleneck method
• Manifold learning
• Vector quantization

Reinforcement learning

• Deterministic policy gradient
• Learning automata
• Proximal policy optimization
• Q-learning
• Soft actor-critic
• State–action–reward–state–action
• Temporal difference learning
• Trust region policy Optimization

Other

• Bayesian belief network
• Bayesian knowledge base
• Deep belief networks
• Deep boltzmann machines
• Deep neural networks
• Discrepancy modelling
• Gaussian naive bayes
• Generative adversarial network
• Hierarchical temporal memory
• Knowledge-enhanced machine learning
• Markov models
• Multinomial naive bayes
• Neural style transfer
• Physics-informed machine learning
• Sparse identification of nonlinear dynamics
• Transformer