Downlink user selection and resource allocation for semi-elastic flows in an OFDM cell

Abstract

We are concerned with user selection and resource allocation in wireless networks for semi-elastic applications such as video conferencing. While many packet scheduling algorithms have been proposed for elastic applications, and many user selection algorithms have been proposed for inelastic applications, little is known about optimal user selection and resource allocation for semielastic applications in wireless networks. We consider user selection and allocation of downlink transmission power and subcarriers in an orthogonal frequency division multiplexing cellular system. We pose a utility maximization problem, but find that direct solution is computationally intractable. We first propose a method that makes joint decisions about user selection and resource allocation by transforming the utility function into a concave function so that convex optimization techniques can be used, resulting in a complexity polynomial in the number of users with a bounded duality gap. This method can be implemented if the network communicates a shadow price for power to power allocation modules, which in turn communicate shadow prices for rate to individual users. We then propose a method that makes separate decisions about user selection and resource allocation, resulting in a complexity linear in the number of users.

Resource Allocation for Semi-Elastic Applications With Outage Constraints in Cellular Networks

“…We are thus motivated to solve a dual problem. The idea, used previously for strictly concave utility functions [7][37] [38], is to separate the determination of each user's average rate (S k,t ) and the allocation of instantaneous power (p) using a set of intermediate variables d = {d k,t , ∀k, t} as bounds on the achieved rates. This decomposition can be used to allow the instantaneous power to be determined on a faster time scale than average rate.…”

“…utility is convex at rates less than a threshold and concave at rates above that threshold [2]. A number of papers model utility as a function of instantaneous rate, and thus propose resource allocation algorithms based on instantaneous rate [3][4] [5][6] [7][8] [9] [10]. For instance, in [7], we considered the case in which utility is a semi-elastic function of the rate achieved in each time slot.…”

“…A number of papers model utility as a function of instantaneous rate, and thus propose resource allocation algorithms based on instantaneous rate [3][4] [5][6] [7][8] [9] [10]. For instance, in [7], we considered the case in which utility is a semi-elastic function of the rate achieved in each time slot. We proposed a near-optimal algorithm that iteratively finds optimal shadow prices for power and rates, and uses these shadow prices to allocate power and subcarriers.…”

“…MPEG, utility should be represented as a sigmoid function of the average rate over each group-of-pictures [11]. Thus the algorithms in [3][4] [5][6] [7] are not suitable to video conferencing. Since a group-ofpictures comprises many time slots, resource allocation based on average rate over each group-of-pictures can exercise considerably greater flexibility than resource allocation based on instantaneous rate, and thereby achieve better results.…”

“…This calls for an approach to resource allocation that considers both utility and outage based on the average rate over the groupof-pictures. None of the prior literature on resource allocation for video using sigmoid utility functions, including our prior papers [7] and [11], considered outage. The existence of an outage constraint poses a challenge to resource allocation.…”