Self-healing and distributed proxy deployment utilizing a decentralized network and reinforcement learning

A method, comprising:

• obtaining, by a system comprising at least one processor, a status of a network comprising a group of proxy nodes that communicate with respective proxy nodes of the group of proxy nodes in a peer-to-peer manner, wherein the status represents a current state of the network with respect to respective point-to-point (P2P) communication links between the respective proxy nodes;
• and
• based on the status of the network, utilizing, by the system, a Q-learning-based reinforcement learning process to facilitate optimization of a future state of the network by modifying, by the system, at least one of
• a proxy selection comprising a selection of proxy nodes to be included in the group of proxy nodes,
• a distribution of at least one of the selection of proxy nodes or a group of virtual proxy nodes within a network topology of the network, or
• a load balancing of network traffic, corresponding to the respective P2P communication links, among the distribution of the at least one of the selection of proxy nodes or the group of virtual proxy nodes, wherein the Q-learning-based reinforcement learning process utilizes respective e-greedy-based decision making strategies to select, during respective states of the network via a Q-table based on respective epsilon-values that are respective defined hyperparameters of the system, respective groups of actions to be performed via the optimization.

A non-transitory machine-readable medium comprising instructions that, in response to execution, cause a system comprising at least one processor to perform operations, the operations comprising:

• receiving information representing a present network state of a network corresponding to respective point-to-point (P2P) communication links between respective proxy nodes of a group of proxy nodes of the network, wherein the proxy nodes are communicatively coupled to the respective proxy nodes in a peer-to-peer manner;
• and
• based on the present network state, employing a Q-learning-based reinforcement learning process to facilitate optimization of a future network state of the network, wherein the optimization comprises modifying at least one of
• a selection of proxy nodes to be included in the group of proxy nodes,
• a distribution of the selection of proxy nodes within a network topology of the network, or
• a load balancing of network traffic among the selection of proxy nodes, and wherein the Q-learning-based reinforcement learning process utilizes respective e-greedy-based decision making strategies to select, during respective states of the network via a Q-table based on respective epsilon-values that are respective defined hyperparameters of the system, respective groups of actions to be performed via the optimization.