Cut-in maneuver recognition and behavior generation using Bayesian networks and fuzzy logic

In this paper a method for recognizing the cutin maneuver using a Bayesian network is proposed, followed by an evaluation of the degree of danger of the situation. An adequate behavior of the ego-vehicle is generated using a fuzzy logic controller. The task of this controller is to provide appropriate actions such as to help prevent a possible collision caused by a dangerous cut-in maneuver, by issuing an emergency brake, as well as to assure a high degree of comfort to the vehicle passengers, i.e. to avoid triggering false emergency brakes. The fuzzy controller provides qualitative and quantitative values, for a HMI and possible actuators, respectively. The qualitative values provide the appropriate response of the ego-vehicle for avoiding an imminent collision, in the form of passive warnings. The quantitative values are the adequate brake values, which should be used by actuators, in order to avoid a collision avoidance maneuver. The solution is based on the environment information provided by a stereovision perception system. The experiments show the solution is able to reason and make decision using the uncertain sensorial data, providing an overall good performance.