A Fundamental Perspective to the Design of Delay-limited Wireless Communications

摘要

The explosive growth of wireless multimedia services has brought severe challenges to the design of cost-effective communications systems. In the future-generation wireless systems, ubiquitous delivery of multimedia data with guaranteed quality-of-service (QoS) is anticipated whereas the physical vulnerability in a wireless channel such as fading continues to be a critical concern. Previous efforts tended to design wireless systems with maximal capacity by, for instance, an opportunistic transmission that adapts its rate according to the channel strength. Such designs are, however, of little relevance for providing real-time delay-sensitive traffic because it would be infeasible to postpone the transmission until the channel improves since the data may expire. In this paper, it is proposed to model a delay-limited communication by a sequence of block (or packet) transmissions and the delay constraint is novelly characterised by the probability that a target rate is not attained for a given number of blocks. Under this setting, the cost of transmission in terms of power is minimised by intelligent power control across the blocks with consideration of both the channel strength and the probabilistic delay constraint. This radical approach corrects the wireless systems design fundamentally to fit with multimedia data traffic. In particular, this paper develops an algorithm to minimise the overall transmit power subject to a given probabilistic delay constraint by adaptive power allocation. Intriguingly it is proved that a 'hard' delay constraint (ie, with zero probability of target being not satisfied) will disable time diversity or the optimal power control will be to pour all the power for the first block and not invest any power for the blocks in the future. Numerical results also demonstrate how the cost of transmission is affected by the various delay-constraint parameters such as rate, number of blocks and outage probability, etc.