GENETIC ALGORITHM
FOR OPTIMIZING
SHORTEST PATH
PROBLEMS IN NETWORK
ROUTING
Amarjit Singh Dhillon Ranjit Singh Saini Karan seth Rokibul Hassan
DATA NETWORKS ROUTING
• Routing: Selecting minimum cost/distance paths for transferring packets from Source node to destination node
• Routing Is a complex process in large networks
• Two Types1. Static Routing2. Dynamic Routing
Source
Destination
Destination
PLANNING PROBLEM
• In static routing algorithms, the path used is fixed regardless of traffic conditions.
Problems :
1. No- Fault tolerance : If failure occurs, traffic will not be re-routed.
2. Time consuming for large network configuration: Manual configuration of each router takes a lot of time.
3. Prone to Human errors: As configuration is done manually.
SHORTEST PATH PLANNING PROCESS
• Finding a minimum cost path between a given number of nodes.
• Existing solutions include Dijkstra algorithm: used in OSPF
A. Inefficient for very large networks.B. A lot of computations need to be repeated for large networks.
O(|E| +|V|log|V|) – optimal method / best soluton
Moreover we are finding paths between two routers s.t all given router are to be covered – using meta-heuristic method
Dijkstra algorithm
INITIAL PARAMETERS - MODIFIED GENETIC ALGORITHM
Input Parameters1. Initial # of routers to be installed2. Initial population size3. # of generations
Outputs 4. 3-D graph showing initial positions of routers/ switches/ nodes.5. 3-D graph showing shortest path between initial and final path.6. 2-d graph showing shortest distance at all generations.
MGA ALGORITHM
a) Randomly generate 3-d co-ordinates for routers/ switches/ nodes.b) Randomly Generate population of initial routes.c) Generate distance – cost matrix; o For 1 : all generations {
a) Update distance – cost matrix among all nodes.b) Compute the fitness for each individual – i.e minimum value of cost;
c) Perform the selection Find Elite Individual on basis of min cost - best individual directly moved to next generation
d) DO Crossover e) if PC ( fixed ) > PRC (random ) { bifurcate initial pop into half and perform crossover }f) Else { skip Crossover }
g) Generate an selection vector to select child populationh) Select individuals in pair of 4 using selection vector and do mutation {
i) Do Mutation -j) if PM ( fixed ) > PRM (random ) { Swapping, Sliding, Flipping }k) Else { skip mutation } }
l) Initial population = new population } END
GENERATING POPULATION
DISTANCE COST MATRIX
RANDOM CROSSOVER / MUTATION PROBABILITY
CROSSOVER
USING SELECTION VECTOR
TWO POINT MUTATION - FLIPPING
SWAPPING & SLIDING
EXPERIMENT RESULTS
CONCLUSION
Provides solution in reasonable amount of time– using meta-heuristic method
Achieves global minima - for given # of generations
Mutation & crossover led to faster convergence
Multiple constraints can be implemented in MGA
REFERENCES
[1] R Kumar and M Kumar, “Exploring Genetic algorithm for shortest path optimization in data networks,” Global Journal of Computer Science, 2010.
[2] C. Ahn and R. S. Ramakrishna, “A genetic algorithm for shortest path routing problem and the sizing of populations,”Evolutionary Computation, IEEE Transactions on, vol. 6, no. 6, pp. 566–579, 2002.
[3] G. T. Nair and K. Sooda, “An intelligent routing approach using genetic algorithms for quality graded network,” International Journal of Intelligent Systems Technologies and Applications, vol. 11, no. 3–4, pp. 196–211, 2012.
[4] J Lee and J Yang, “A Fast and Scalable Re-routing Algorithm based on Shortest Path and Genetic Algorithms J. Lee, J. Yang Jungkyu Lee,” International Journal of Computers Communications & …, 2014.
CONCLUSION