Estimation of Ad Hoc Communication Channel using Stochastic Filters

Ad hoc networks differ from conventional cellular networks in terms of their flexible infrastructure. In a cellular network, either the transmitter or the receiver is stationary, whereas in ad hoc networks, there is direct communication between mobile transmitters and receivers via a wireless medium. The temporal variations occurring in the propagation environment of ad hoc networks are mainly due to the nodes moving at varying speeds and the instantaneous appearance or disappearance of existing paths between transmitters and receivers. The time-varying statistics of wireless channels are known to be captured effectively by the dynamic models based on stochastic differential equations (SDEs). This paper proposes a scheme for the estimation/ identification of an ad hoc channel dynamically modelled via SDEs using the Ensemble square root filter (EnSRF). The channel states and parameters are estimated from the dynamic models arrived at by the factorization of the fundamental channel characteristic called Doppler power spectral density (DPSD), followed by a stochastic realization. The stochastic realization brings the channel model to a state space form viable for the chosen stochastic filtering frameworks. The constant parameters of the channel model, so derived, are realized using random walks and are augmented to the state space to enable a joint state and parameter estimation. A dynamic SDE-based channel model belonging to an urban environment with a predominant fading is adopted in numerical illustrations. Simulation results indicate remarkably superior accuracy and filter convergence of EnSRF estimates.