Reducing Congestion Effects by Multipath Routing in Wireless Networks

We propose a solution to improve fairness and increasethroughput in wireless networks with location information.Our approach consists of a multipath routing protocol, BiasedGeographical Routing (BGR), and two congestion controlalgorithms, In-Network Packet Scatter (IPS) and End-to-EndPacket Scatter (EPS), which leverage BGR to avoid the congestedareas of the network. BGR achieves good performancewhile incurring a communication overhead of just 1 byte perdata packet, and has a computational complexity similar togreedy geographic routing. IPS alleviates transient congestion bysplitting traffic immediately before the congested areas. In contrast,EPS alleviates long term congestion by splitting the flow atthe source, and performing rate control. EPS selects the pathsdynamically, and uses a less aggressive congestion controlmechanism on non-greedy paths to improve energy efficiency.Simulation and experimental results show that our solutionachieves its objectives. Extensive ns-2 simulations show that oursolution improves both fairness and throughput as compared tosingle path greedy routing. Our solution reduces the variance ofthroughput across all flows by 35%, reduction which is mainlyachieved by increasing throughput of long-range flows witharound 70%. Furthermore, overall network throughput increasesby approximately 10%. Experimental results on a 50-node testbed are consistent with our simulation results, suggestingthat BGR is effective in practice.