Abstract. When a node is moving in a wireless network, the routing metrics associated to its wireless links may reflect link quality degradations and help the routing process to adapt its routes. Unfortunately, an important delay between the metric estimation and its inclusion in the routing process makes this approach inefficient. In this paper, we introduce an algorithm that predicts metric values a few seconds in advance, in order to compensate the delay involved by the link quality measurement and their dissemination by the routing protocol. We consider classical metrics, in particular ETX (Expected Transmission Count) and ETT (Expected Transmission Time), but we combine their computations to our prediction algorithm. Extensive simulations show the route enhancement as the Packet Delivery Ratio (PDR) is close to 1 in presence of mobility.

International audienceIn wireless mobile ad hoc networks, routing protocols use metrics to select the best routes. Metrics may reflect the quality of the wireless link and help to manage mobility. But, there is a delay between link quality measurements performed by the routing protocol and their instantaneous values. To compensate this delay, the idea is to anticipate the value of the metric a few seconds in advance. The purpose of this paper is to propose a new technique to compute ETX metric, that is sensitive to mobility and which optimizes throughput at the same time. We show through simulations performed with NS-3 (Network Simulator version 3) that our approach leads to Packet Delivery Ratio (PDR) close to 1 in presence of mobility

In Mobile ad-hoc Networks (MANETs), node mobility management is performed by the routing protocol. It may use metrics to reflect links state/quality. But, the delay between measures of the link quality and its integration in the route computation is very detrimental to the mobility management. Consequently, routing protocols may use lossy links for a few seconds leading to a significant performance deterioration. In this paper, we propose a new routing metric technique calculation which aims at anticipating link quality. Basically, the idea is to predict metric values a few seconds in advance, in order to compensate the delay involved by the link quality measurement and their dissemination by the routing protocol. Our technique is based on measurements of signal strength and is integrated in two classical routing metrics: ETX (Expected Transmission Count) and ETT (Expected Transmission Time). Validations are performed through both simulations and a testbed experimentation with OLSR as routing protocol. NS-3 simulations show that our metric may lead to a perfect mobility management with a packet delivery ratio of 100%. Experiments on a testbed prove the feasibility of our approach and show that this technique reduces the packet error rate by a factor of 3 in an indoor environment compared to the classical metrics calculation.