Abstract:

An adaptive heuristic critic (AHC) Reinforcement Learning algorithm is presented for the dynamic spectrum allocation in an autonomously deciding mode in heterogeneous radio networks based on the normalized radial basis function (NRBF). The algorithm accelerates the learning speed by utilizing the NRBF when constructing the state space, and improves the learning efficiency by using the AHC scheme to reduce the unnecessary exploration. Through interactions with the radio environment, it learns to allocate the proper frequency band for each session in multiple radio access networks. Simulation results show that the proposed algorithm can lead to a better spectrum efficiency and quality of service compared with to the fixed frequency planning scheme or general dynamic spectrum allocation policy.

Key words: heterogeneous radio networks, dynamic spectrum allocation, reinforcement learning, normalized radial basis function