Modelling a Decentralized Constraint Satisfaction Solver for Collision-Free Channel Access
In this paper, the problem of assigning channel slots to a number of contending stations is modeled as a Constraint Satisfaction Problem (CSP). A learning MAC protocol that uses deterministic backoffs after successful transmissions is used as a decentralized solver for the CSP. The convergence process of the solver is modeled by an absorbing Markov Chain (MC), and analytical, closed-form expressions for its transition probabilities are derived. Using these, the expected number of steps required to reach a solution is found. The analysis is validated by means of simulations and the model is extended to account for the presence of channel errors. The results are applicable in various resource allocation scenarios in wireless networks.