Real-time Detection of Pathological Traffic Situations via AIS

This paper suggests using AIS (Artificial Immune Systems) for automatically detecting and analyzing traffic anomalies due to unpredictable road situations. Various authors [DeCaTi2002], [ForHofSom1996], [TarSkorSoko2003] suggested models and methodologies based on the paradigm of the biological immune system in the framework of network intrusion detection [MykHebLev1994]. We suggest applying an AIS-based recognition model for detecting and controlling pathological road situations, such as traffic jams due to unexpected events. Our approach is based on the idea of defining pathological traffic patterns as antigens, and applying the singular value decomposition (SVD) of matrices of previous abnormal traffic logs to compute matching antibodies. Such antibodies will subsequently be applied to recognize actual traffic patterns as either self (normal traffic), or non-self (pathological traffic). This will be done by (a) using antibodies to map past pathological traffic logs into a subset of a k-dimensional shape space, and cluster them into recognition balls, (b) mapping every new incoming traffic log into a point P of the same shape space, and (c) determine the nature of the incoming traffic log by evaluating its minimum distance recognition balls.