NEFCOP: A Neuro-Fuzzy Vehicle Collision Prediction System
K. Venkatesh, Aduthotla Ramesh, M. Alagusundaram, J.V. Sahana · 2006
Given that road accidents occur in a real-time environment, simple crisp functions would barely provide an estimate of the gravity of the life situation. Fuzzy-based systems are able to establish complex non-linear relationships between variables with ease, making them perfect in the domain of vehicle collision prediction. However a sole fuzzy-based system, would fail to provide the user-specificity and intuition needed here. Neural networks, through their intelligent learning capabilities are ideal for this requirement. We propose NEFCOP, a neuro-fuzzy vehicle collision prediction system. NEFCOP uses laser ranging to obtain information about the road environment, which is then passed to a two-stage prediction system. The first stage clusters these data in order to prioritize them based on their relevance. The second stage is a neuro-fuzzy sub-system, which processes these data analyzing the possibility of a collision, following which the driver is warned accordingly. Thus, NEFCOP achieves adaptive, realistic accident prediction.