MDP based link scheduling in wireless networks to maximize the reliability

摘要

This paper considers wireless networks where communication links are unstable and link interference is a challenge to design high performance scheduling algorithms. Wireless links are time varying and are modeled by Markov stochastic processes. The problem of designing an optimal link scheduling algorithm to maximize the expected reliability of the network is formulated into a Markov Decision Process first. The optimal solution can be obtained by the finite backward induction algorithm. However, the time complexity is very high. Thus, we develop an approximate link scheduling algorithm with an approximate ratio of 2 (N - 1) (r(M)Delta - r(m)delta); where N is the number of decision epochs, r(M) is the maximum link reliability, r(m) is the minimum link reliability, Delta is the number of links in the largest maximal independent set and delta is the number of links in the smallest maximal independent set. Simulations are conducted in different scenarios under different network topologies.