the simulation and analysis of OFDMA subcarrier allocation techniques in multicellular environments. the performance evaluation of simple algorithms compared to a more sophisticated and computationally demanding scheme.
Base Station selection the best server of each user is determined by the channel gain at the (system s) central frequency. Subcarrier Allocation (SA) Determination of the number of subcarriers (s u ). Assignment of the specific subcarriers to users. Bit Loading and Power control Definition of the transmit power of the user and modifications on transmit power of co-channel users in order to preserve QoS.
Sequential BSs allocate subcarriers sequentially. The new user is allocated the first s u available carriers of his cell. Coordinated The network cells are divided into two groups. Random 1 st group 2 nd group 1 2 N The BS selects randomly carriers from the set of the available ones. Random-Coordinated Cell is split into two regions, internal and external, with a gain threshold criterion Internal region employs random subcarrier allocation External region employs coordinated subcarrier allocation 4-Coord Cell splits into 4 regions with a gain threshold criterion. Available channels are also divided into 4 groups. Internal regions employs random SA. Coordination is applied for the 3 regions
Adaptive The BS re-allocates all the subcarriers of the cell to its users taking into account CSI (channel gain + interference). Each user is allocated the subcarrier with the best normalized channel gain. G u,n,k (I u,n,k + N 0 )SNR GAP Priority is given to the user who first entered the network.
System Parameter Number of Cells 7 Radius Central Frequency Available Bandwidth Simulation value 0.8Km f c =2.5GHz 10MHz Number of Subcarriers N=128 Subcarrier Spacing BER 10-5 Modulation Service Sets f = 78.125Kbps QPSK (single) QPSK-16QAM-64QAM (adaptive) 128-256-512Kbps (low rate) 512-1024-2048Kbps (high rate)
The simulations employed a hybrid approach for the propagation model for the radio channel: characteristics for the central cell were produced with the help of ray tracing, while for the rest of the cells they were produced with the Okumura- Hata pathloss model. The coherence bandwidth, calculated from the RMS delay spread of the predicted impulse response at each point of the area (grid resolution 10m) showed that for almost all cases in the studied operational environment, the coherence bandwidth is higher than 100KHz. This has led to the production of the relevant impulse responses for each of the employed 128 channels (5MHz around f o ).
During each iteration: Users enter the network serially and uniformly distributed. The service requirement of each user is randomly selected. a user is rejected when the network fails to satisfy the SINR requirements. The presented results have been averaged over 300 Monte Carlo simulation and concern the central cell. The execution time was reduced to feasible levels through the development of a grid- enabled problem solving environment.
The adaptive RRM outperforms the other four schemes. The adaptive RRM retains an almost stable LF while the first four schemes increase their LF for higher BL probabilities. Random carrier allocation presents the second best behavior.
for higher services the coordinated approach improves its performance, while the random scheme shows the opposite behaviour the advantage of adaptive algorithm is diminished in higher BL probabilities (>30%)
adaptive modulation enhance the performance of all schemes (in particular of random). the difference between the adaptive and the random is decreasing the adaptive loses its advantage when BL>20%
the coordinated algorithm fails to retain its advantage in higher data rates. It is clearly outperformed by the random algorithm. the randomcoordinated algorithm shows a worth mentioning performance only in high BL probabilities (>40% which is not acceptable in real network deployments).
The random algorithm deteriorates its performance in higher service sets when fixed modulation is applied. The coordinated is clearly more suitable in that case. Adaptive modulation adds 9Mbps in the performance of the random algorithm (1.16 10.40) Fixed Modulation Set 1 Coord Rand Adapt P BL =5% 0.6 1.16 6.01 P BL =20% 3.89 4.58 7.10 Set 2 Coord Rand Adapt P BL =5% 1.98 1.00 5.77 P BL =20% 4.83 1.92 7.61 Adaptive Modulation Set 1 Coord Rand Adapt P BL =5% 0.72 4.00 8.68 P BL =20% 4.28 10.40 9.57 Set 2 Coord Rand Adapt P BL =5% 2.39 5.69 10.38 P BL =20% 5.90 11.37 13.13
Based on the concept of R-C The cell area is now divided into 4 co-centric with equal radii (uniform distribution for the users location). The coordination of areas 2-4 is adaptively evaluated according to 1 2 3 4 the setup of each iteration, so that the interference levels are reduced.
fixed user channel: n CH = 4 fixed modulation: QPSK Users tried to enter the network: 100