Routing in a Sparse Graph: a Distributed Q-Learning Approach

A distributed Q-learning approach is proposed for solving the problem of routing messages through a sparse graph with minimal cost. Each node in the graph functions as an independent agent, learning how to forward messages effectively based on the final target node. Agents update their individual Q-tables based on path costs and use an epsilon-greedy strategy to balance exploration with exploiting the best-known routes. The article provides the mathematical formulation for the Q-learning update rule and Python code for implementation.