Topology Estimation in Mobile Ad Hoc Networks Master Thesis

The term ad hoc indicates the capability that two or several nodes can connect and communicate with each other independently of a centralized infrastructure. The disclaimer of such a central unit yields to the requirement of a self-organizing behavior of the network where the nodes organize themselves in a distributed manner. Knowing the geographical representation of the network can help on the organization of such ad hoc networks. The location of a node is a very helpful information and can be used by routing and service provisioning. Based on the previous or recent locations, the nodes future location can be predicted which provides a continuous and reliable working of these tasks. This leads to the task of mobility prediction where the location of the nodes must be known for predicting the future locations. In this thesis, we propose an approach to the estimation of a node’s near topology in mobile ad hoc networks. First, each node estimates its distance to all its one-hop neighbors using the received signal strength (RSS). These measured RSS values are mapped to distances. Second, each nodes estimates its distance to all its two-hop neighbors based on the one-hop distances. Third, we apply MDS techniques to reduce the representation space from Rn to R2 which provides a representation of the near topology in a 2-D Euclidean space. Eventually, we propose different feedback mechanisms to improve the performance of each of the previous steps. We evaluate our approach in a testbed and by simulations using the CSG testbed and Ns-2, respectively. The simulations show that the feedback mechanisms provide a fast improvement and reduces the error by 32.7%. This is the result of a real testbed using real measurements.