Applying location estimation for reliable routing in tactical unmanned ground vehicle networks

Abstract

An unmanned ground vehicle (UGV) network operates as an ad hoc autonomous tactical network. In this paper, we study the deployment of a UGV network and its ability to use location estimation to provide reliable data forwarding. This analysis involves (1) location estimation to obtain realistic estimates of the node’s (UGV) spatial positions and (2) real-time path planning to obtain reliable routing among nodes using the estimates of the node positions. Conventional approaches commonly use the popular Random Waypoint (RWP) model (or its variants) as the mobility model. The RWP approach uses randomly selected node positions and has no way to cope with nodes moving in and out of transmission range. These deficiencies cause the route reliability to suffer. Our approach (1) uses an Extended Kalman Filter (EKF) for node mobility/location estimation and (2) uses a modified Ad Hoc On Demand Distance Vector (AODV) algorithm for route selection using a new routing parameter called contact time. We refer to this combined approach as the AODV-LocPred algorithm. We compare the AODV-LocPred algorithm with AODV using the RWP model. This is referred to as AODV-RWP. Our simulations show that the AODV-LocPred algorithm significantly outperforms the AODV-RWP algorithm in terms of packet delivery ratio (PDR), because it provides realistic node position estimates and it copes in a reasonable way with nodes moving in and out of transmission range. However, the AODV-LocPred incurs an increase in end-to-end delay. Nonetheless, we expect that in many real-world settings, this tradeoff is acceptable.