Topology control in mobile ad-hoc networks allows better spatial reuse of the wireless channel and control over network resources. Topology control algorithms tend to optimize network power usage by keeping the topology connected. However, few efforts have focused on the issue of topology control with mobility. One of the most efficient mobility aware topology control protocols is the "mobility aware distributed topology control protocol". The major problem with this protocol is the future distance predictor which uses mobility prediction to estimate the future distance of neighboring nodes. The efficiency of this estimator varies in presence of different mobility models, sampling rates and different speed ranges. In this paper, we introduce an adaptive mobility prediction method that uses learning automaton to estimate the coefficients of a simple adaptive filter in order to predict the future distance of two neighboring nodes. We evaluated this estimator in the mobility aware distributed topology control protocol. Simulation results show significant improvement in accuracy of the future distance prediction and reduction in power consumption of each node.