Vision-based active safety system for automatic stopping

Intelligent systems designed to reduce highway fatalities have been widely applied in the automotive sector in the last decade. Of all users of transport systems, pedestrians are the most vulnerable in crashes as they are unprotected. This paper deals with an autonomous intelligent emergency system designed to avoid collisions with pedestrians. The system consists of a fuzzy controller based on the time-to-collision estimate – obtained via a vision-based system – and the wheel-locking probability – obtained via the vehicle’s CAN bus – that generates a safe braking action. The system has been tested in a real car – a convertible Citroën C3 Pluriel – equipped with an automated electro-hydraulic braking system capable of working in parallel with the vehicle’s original braking circuit. The system is used as a last resort in the case that an unexpected pedestrian is in the lane and all the warnings have failed to produce a response from the driver.

► We have designed a last resort system to avoid collisions with pedestrians.
► Time-to-collision to unexpected pedestrians is estimated via stereovision.
► Wheel-locking probability is also used to generate a soft control action.
► Fuzzy logic combines both inputs to calculate the automatic maneuver.
► Real tests in a commercial car demonstrate the feasibility of the system.