Design and analysis of downlink utility-based schedulers

We consider scheduling data tra c for the downlink of a wireless network A draining problem is formulated where the goal is to transmit a given set of packets Each packet is assigned a utility that depends on the delay incurred We propose a simple gradient based scheduling rule which attempts to maximize the average utility per packet A deterministic analysis of this rule is given by considering an asymptotic uid limit where the number of packets becomes large while the packet size decreases to zero In this limiting regime we formulate an optimal control problem which corresponds to nding the best scheduling policy Using Pontryagin s minimum principle we prove that in a special case the gradient based algorithm is optimal Simulations are presented to illustrate these results