110 likes | 220 Views
Efficient Routing via Clustering Strategy for Ad Hoc Networks. by: Gaurav Chopra (00D07013) Satyam Srivastava (00D07014). Problem Formulation. Aim of the Project Clusters formed should be spatio-temporal stable Build Efficient Routing Algorithm using the cluster framework
E N D
Efficient Routing via Clustering Strategy for Ad Hoc Networks by: Gaurav Chopra (00D07013) Satyam Srivastava (00D07014)
Problem Formulation • Aim of the Project • Clusters formed should be spatio-temporal stable • Build Efficient Routing Algorithm using the cluster framework • Compare its performance with existing clustering strategy • Constraints • Cluster Head • Should be located centrally for spatial stability • Should have high energy value relatively to other nodes in the cluster • Should have a history of being less mobile • Cluster Size • Overlap between clusters should be minimum • Should suffice the connectivity constraints • Has a maximum size (number of member) constraint
System Model • Cluster Formation: When a node switches ON, it tries to join/form a cluster using following procedure: • Transmits a hello packet (Broadcast Packet) and waits for a reply for a fixed amount of time (reply_timeout). • If any of the cluster head is around and it listens to this hello packet, it replies to the request. • If the node does not receive any reply before reply_timeout, it will assume that no clusters are present around and will become the cluster head itself. • If it receives a reply it will send an ACK to the cluster head which replied. • If it receives more than reply it will look for the one with the highest SINR value and will send an ACK to that cluster head. • The cluster head after receiving the ACK, informs the newly joined node about the cluster. The cluster head informs the node about the cluster member’s id and their location. • The cluster head sends an update packet containing the information about the new member to inform other members about the event.
System Model (contd.) • Cluster Head (CH) selection from proposedVoting Algorithm • Each of the cluster members has to periodically votes for the best location of the CH for itself. • Each node evaluates the best possible position using the location of the cluster members and a weight function wi,j • Present CH receives the votes from all the nodes in the cluster and evaluate their mean to estimate the optimal position for the new possible CH.
System Model (contd.) • Cluster Maintenance: Mobility of nodes introduces the need for cluster division, cluster fusion and handoff. • Cluster Fusion: • A cluster completely moves inside another cluster. • Around 90% of the cluster nodes are inside another cluster • Cluster Division: • Number of the members in the cluster goes aboveMax_Num_Cluster_Members • Handoff • Soft Handoff • Hard Handoff
Routing Strategy • Standard protocols are used to test the effectiveness of the Novel Clustering technique proposed • Intra Cluster • DSDV – Destination Sequenced Distance Vector • Inter Cluster • AODV – Ad-hoc On-demand Distance Vector • Arrival Process • Poisson generated packets for varying load per node (0.01 to 0.25)
Simulation – Models Used • Network • X and Y coordinates of the nodes are generated by uniformly distributed random variable • Mobility Models • initial directions in degree are generated by the uniform distribution U[0,360] • After negative exponentially distributed amount of time, with mean equal to 60 seconds, the direction of a user is changed • The new direction is generated by a Gaussian distribution with mean equal to the user's old direction and standard deviation 30 degrees • Velocity of nodes are fixed • Channel Model • log-normal shadowing model
Conclusion • Throughput performance: From the plots throughput performance is almost same in both the strategies. Our strategy performs better at low load but performance coincides as load increases. • Power performance: Our approach surpasses fixed rectangular clustering strategy when number of clusters are less (9 and 12) but performance deteriorate as number of cluster increases to very high value (25). Performances of both strategies match when number of clusters was 16. • Stability of Cluster Framework • Cluster Head located centrally • Spatio-temporal stability • Sufficed Connectivity constraints • Thus, Novel Clustering performs much better and is suitable for a network having high node density.