Fusion of sensors applied to a collision avoidance system

This article describes a multisensor system developed in order to avoid road crashes. The system deals with data issued from sensors placed on an experimental vehicle in a real environment. Indeed, actually no technology allows to conceive an ideal sensor responding to the safety conditions in the different situations met in road driving. Our objective consists in establishing a clever fusion of data issued from different existing sensors. The experimental vehicle is equipped with different sensors (a microwave anti-collision radar, a laser telemeter and a linear stereo vision system). The goal of this project is to design a hardware and software system which can prevent road crashes. The integration of the multisensor system aims at fusing different information sources in order to take safer decisions. A lot of heterogeneous and asynchronous data must be taken into consideration for the study and achievement of an on-board driving aid. Information given by different sensors pass through a Controller Area Network. The first step is the research of a common spatio-temporal referential for the different data. The sensors limitations are taken into account and the system takes advantage from each sensor. The sensor set allows the reconstruction of a local road map as accurately as possible. Copyright (C) 1998 IFAC.