Bandwidth Allocation for Handover calls in Mobile Wireless Cellular Networks –
Genetic Algorithm Approach
Khaja Kamaluddin, Abdalla Radwan [email protected], [email protected]
Computer Science Department Faculty of Science
Sirte University Libya
Khaja.KSirte University, Libya ACIT – 2010
Acknowledgements
We are very much thankful to the management of Sirte University, Libya for supporting and facilitating in this research work.
Our sincere thanks to reviewers for providing their valuable comments and suggestions.
Khaja.KSirte University, Libya ACIT – 2010
Objectives
Channel allocation using Genetic Algorithm Channel allocation based upon fitness score Minimum allocation in worst case. Maximum bandwidth utilisation Avoiding wastage of cell bandwidth.
Khaja.KSirte University, Libya ACIT – 2010
Problems
Cell size is being reduced Frequent handovers take place Demand for wireless connectivity is increased Available resources are limited Increase of Blocking/Dropping probability
Khaja.KSirte University, Libya ACIT – 2010
Existing solutions
1. Using Guard channels
2. Centralized channel allocation
3. Distributed dynamic channel allocation
4. Co operative non cooperative resource allocation
5. Proper utilization of available bandwidth
6. Utilization by accurate prediction
7. Online load balancing
8. Increase system utilization with degradation of QOS.
Khaja.KSirte University, Libya ACIT – 2010
Problems with existing solutions
Reduced dropping probability but wastage of resources in absence of calls
If central system fails whole network is in problem Mobile tracking and prediction is always may not be
correct. Improper utilization of bandwidth Degradation of QOS Channels exhausted
Khaja.KSirte University, Libya ACIT – 2010
Solution
Channel Allocation
by
Genetic Algorithm
Fitness Function
Population
Selection
Crossover
Mutation
New Population
Figure1. Genetic Algorithm
Khaja.KSirte University, Libya ACIT – 2010
Proposed System ModelGenerate population
Create random handover mobile nodes/calls and random time slots
Evaluate the fitness
Previously used time slot duration full or partial or slot not used
Selection Random number generation, assignment and ascending order.
Crossover Node + time slot
Mutation Change in time slot duration
Elitism Allocate Requested
New population Time slot allocated nodes, empty slots if any
Defined Details Fitness Score
11 Fully utilized 3
10 Partially utilized 2
01 Not utilized 1
00 Bottlenecked 0
Khaja.KSirte University, Libya ACIT – 2010
Proposed Solution
Bandwidth allocation – GE Approach Population of chromosomes – Handover calls &
Time slots Genes – Bandwidth requirement (Time slots) Fitness Value – Previous History of the Call
Khaja.KSirte University, Libya ACIT – 2010
Genetic Bandwidth Allocation
Initialise Population
Crossover
Fitness Function
Elitism
Selection
Mutation
New Fitness Score
New Population
Discard
Figure2. Modified Genetic Algorithm
Khaja.KSirte University, Libya ACIT – 2010
Crossover OperationM = {m1, m2, m3, …} ---------------- (1)T = {t1, t2, t3, ..} ------------------------(2)B1 = (∑T)/M -----------------------------(3)
Fitness Function EvaluationFitness Score 3 (Fully utilised BW) --- GROUP – I Fitness Score 2 (Partially utilised BW) ---- GROUP – II Fitness Score 1 & 0 ------ Discarded calls
f(Group) = Fitness (Score)
Khaja.KSirte University, Libya ACIT – 2010
Generate of Population
M is set of Randomly generated nodes M = {1m, 2m, 3m, ……..} M = {m│m ε M} T is set of randomly generated time slots T = {t1, t2, t3, ………….} M1 is set of calls with fitness score 3 M2 is set of calls with fitness score 2 M1 ε M and M2 ε M M1 = {m | m is Group1call}. M2 = {m | 0 ≤ Group2 call ≤ t}
Khaja.KSirte University, Libya ACIT – 2010
Fitness score & Calls Arrangement
Fitness score identification
Random number generation
Random number assignment to calls
Arrangement of calls in ascending order based on random number.
Khaja.KSirte University, Libya ACIT – 2010
Selection Process
M1 are allocated as per random number
M2 are allocated as per random number
MutationM1 = Requested allocation
M2 = Requested allocation || Minimum allocation
Khaja.KSirte University, Libya ACIT – 2010
Simulation Scenario
No. of Channels in Cell = 10 IS – 136 TDMA system, Each channel = 6 time slots. Half rate TDMA One slot per frame per customer is dedicated.
Khaja.KSirte University, Libya ACIT – 2010
Simulation Scenario
Total Time slots = T Total Calls = M M = {M1} + {M2} Bandwidth allocation for M1 calls = T1 slots Bandwidth allocation for M2 calls = T2 slots T2 = T – T1 Bandwidth allocation for each M2 call = T2 = (T –
T1)/M2 First interval of time: Randomly generated calls and time
slots. Fixed 0.1 unit of time slot is the minimum bandwidth
Khaja.KSirte University, Libya ACIT – 2010
Analytical ResultsBandwidth Allocation for Handover calls
0
5
10
15
20
t1 t2 t3 t4 t5 t6 t7 t8 t9 t10Time
To
tal C
alls
0
0.05
0.1
0.15
0.2
0.25
Ban
dw
idth
Total G1 Calls Bandw idth allocation to each call
Bandwidth Allocation for all Calls• All handover calls are
accommodated with minimum duration time slot.
Bandwidth Allocation for Lower Fitness Calls
6.46.66.8
77.27.47.67.8
88.2
t1 t2 t3 t4 t5 t6 t7 t8 t9 t10Time
Lo
wer
Fit
ness C
alls
0
0.05
0.1
0.15
0.2
0.25
0.3
Ban
dw
idth
Low er Fitness calls Bandw idthfor LFC
Bandwidth Allocation for LFCs• Assuming that 20% - 30% are higher
fitness value calls and remaining are lower fitness ones.
Khaja.KSirte University, Libya ACIT – 2010
Conclusions Channel Allocation – Genetic Algorithm Time slot Allocation – Fitness score Higher fitness – Priority Lower fitness – Minimum in worst-case Maximum – Bandwidth utilization Efficient – Bandwidth Management Avoided – wastage of cell bandwidth Minimum – call dropping
Khaja.KSirte University, Libya ACIT – 2010
Future Work
We are in the process of evaluating and monitoring the behavior, new fitness function and dropping probability for handover calls, which will be published later.
Khaja.KSirte University, Libya ACIT – 2010
Questions
Khaja.KSirte University, Libya ACIT – 2010
Thank you