Path loss predictions for fringe areas using adaptive neuro-fuzzy inference system

Path loss prediction is an essential technique in wireless radio network designing and development, it supports in perceiving the nature of electromagnetic radio signals in a particularized transmission medium. Varieties of empirical path loss prediction models are available for path loss calculations, but these models are complex and require information regarding environmental conditions for accurate prediction. In this paper, an Adaptive Neuro-Fuzzy Inference System (ANFIS) is explored for the prediction of path loss for multi-transmitter radio wave propagation. A drive test has been conducted for field measurement at Uttarakhand collecting RSSI (Received Signal Strength Indicator) from an individual transmitter. The frequency of transmission of each transmitter is 1800 MHz respectively. An elementary four-layer ANFIS architecture has been optimally trained using RSSI data to accurately map the input values to the equivalent path loss values. The membership function is selected to provide an enhanced and stable mapping of the input to the corresponding output at the lowest number of epochs. The ANFIS model predicts the minimum value of Root Mean Square Error (RMSE) as compared to other path loss models. The obtained ANFIS model also validated a good generalization capability when deployed at a similar terrain. Developed path loss model exhibits desirable qualities for radio network planning i.e. simplicity, accuracy, and better generalization ability which is important for radio coverage prediction.