|M.Sc Student||Tal Uri|
|Subject||Power Diversity for Maximum Delay-Constrained Throughput|
in ALOHA Networks with Successive Decoding
|Department||Department of Electrical and Computer Engineering||Supervisor||ASSOCIATE PROFESSOR Yitzhak Birk|
This work addresses the problem of maximizing the throughput in ALOHA networks subject to a delay constraint (deadline) and a maximum permissible probability of exceeding it. This goal is motivated by one of the main uses of ALOHA at present, namely remote terminals connected via satellite and used for transaction processing.
Increasing the per-message resource expenditure increases the probability of its successful reception prior to the deadline. However, this pollutes the channel and decreases the attainable throughput. By postponing most of the resource expenditure to later transmission rounds, which only occur if the earlier ones fail, the average per-message expenditure, and consequently the capacity loss, can be kept to a minimum. Recently, the judicious use of power diversity in conjunction with CDMA channels and simple matched-filter receivers were considered. The use of power diversity in this case was shown to entail an acute compromise in channel capacity.
In the current work, we again study the judicious use of power diversity over CDMA channels for delay-constrained throughput maximization. However, we assume successive-decoding receivers. For single round transmissions, power diversity is shown to double the capacity relative to power equalization. For multiple transmission rounds, we extend the single-round results and show that the failure probability, even with only two or three transmission rounds, can be diminished almost completely while attaining throughput that is almost equal to the unconstrained one.
Restrictions on maximum permissible mean transmission power limit permissible dynamic range of transmission power and, as a result, reduce the attainable unconstrained throughput. We nonetheless show that our schemes enable the attainment of delay-constrained throughput near the unconstrained one.