Mrcib: Design Of An Efficient Multipath Routing Model For Reducing Network Collisions Via Incremental Bioinspired Optimizations

Abstract

The probability of packet collisions increases exponentially w.r.t. number of communications due to which design of collision-control protocols is of primary importance for large-scale wireless networks. Existing collision control models either showcase high complexity, or have slower response, due to which their efficiency levels are limited for larger networks. Moreover, most of these models cannot be scaled due inherent computational redundancies. To overcome these issues, this text proposes design of a novel multipath routing model for reducing network collisions via incremental bioinspired optimizations. The proposed model initially deploys a Bacterial Foraging Optimizer (BFO) for segregating packets into multipath requests. This segregation is done based on temporal routing performance for different set of paths. The segregated packets are transmitted over the network via an efficient set of multiple paths which are identified by Genetic Algorithm (GA) optimizations. The selected paths are incrementally tuned by a Q-Learning based layer, that assists in reducing collisions under large number of packet requests. This is done via temporal evaluation of network parameters, and continuous model updates for high-efficiency operations. The proposed model was tested under large-scale networks with heterogeneous requests, and it was observed that the model was able to improve communication speed by 8.5%, reduce energy consumption by 6.4%, maintain high packet delivery performance and improve data rate by 3.9% when compared with existing multipath routing protocols that support congestion control for large-scale network scenarios.