Reinforcement learning and CMAC-based adaptive routing for MANETs

A novel routing scheme, which combines the on-demand routing capability of ad hoc on-demand vector (AODV) routing protocol with a Q-routing inspired route selection mechanism, is proposed in this paper. The scheme makes routing choices based on local information (such as mobility, power remaining at the neighbouring nodes) and past experience. The CMAC (cerebellar model articulation controller) function approximator is used to accelerate the reinforcement learning. Through extensive simulation, we demonstrate that our scheme is effective in improving end-to-end delay, without requiring much of the limited network resources.