Table 13.

A categorization of key challenges and potential solutions for fog computing.

Cat *	Challenges	Challenge Description	IRM **	Solution Prospect	IA ***
Computational Resource and Latency	Limited Resources	Processing Power	Complex models undeployable	Resource-efficient algorithms	DL Algorithms (e.g., CNN, RNN)
		Memory		Efficient model training techniques
		Storage		Lightweight models
				Compressed models
	Latency Constraints	Critical Factor for real-time applications	Miss time-critical task	Prioritize time-bond tasks to meet the deadline	DRL
		Network delay can impact performance
	Energy Consumption	Balancing performance and energy-efficient	Increase operation cost	Optimizes resource allocation for critical tasks while considering energy efficiency	SVM
			Reduce device life span
Scalability and Management	Scalability	Dynamic workload	System complexity	Auto Scaling	ML/DL
		Large-scale fog node management	Resource Provisioning	Distributed resource management
	Deployment and maintenance	Complex deployment due to node heterogeneity	Monitoring	Resource allocation	ML/DL (DQN, DDQN, DRL, MORL, ML)
		Infrastructure	Fault Tolerance	Task offloading
				Enhance QoS
	Resource Allocation	Dynamic workload	Cost-effectiveness	Task Scheduling	(ML, DRL, NN, FNN, DQN)
		Resource heterogeneity	User experience	Resource provisioning
		Latency constraints	System performance
Data Availability and Quality	Limited Data	Insufficient historical data is available for training	Inaccurate Model	Generate synthetic data
			Performance inefficiency	Transfer learning
				Data augmentation techniques
	Data Heterogeneity	Data Processing	Inefficient resource utilization	Effective data management
		Data integration	·Model Complexity	Robust ML/DL techniques used
	Data Privacy and Security	Inconsistency format Difficult Data Aggregation	Computation overhead Storage requirement	Federated learning
Model Complexity and Interpretability	Model Complexity	Computation overhead	Potential performance issues	Model optimization	ML/DL
		Overburden		Efficient resource utilization
	Model Explainability	Trust and reliability	decision-making	Recognize patterns	ML/DL
	Model Adaptability	Adopt network conditions	Deployment efficiency	Learning Models	DL
				Value updating

* Categorization of challenges; ** impact on resource management; *** impact algorithm.