Event-Based System for Generation of Traffic Services in Road Congestions - Foundations on Natural and Artificial Computation - page 206

Fig. 4. Generalized fusion model
different sensors extended Kalman filters (EKF) run a set of different kinemat-
ical models as explained in Section 5.2. Thanks to the MEM (Micro-Electro-
Mechanical) technology, low cost inertial sensors can be considered, at the ex-
pense of measurement noises and low level of performance [8]. To communicate
the vehicles in the scene, WLAN ad hoc networks are used [9]. The test vehicle
equipment is a Wireless LAN IEEE 802.11b and antenna by Cisco, EGNOS ca-
pable GPS and DGPS sensors by Novatel and Trimble, and MEM based IMUs
by Crossbow and Xsens.
5.1 Fusion Model
Multiple sources fusion is approached in the current literature from many dif-
ferent points of view. Traditional fusion techniques dedicated to sensor mea-
surements at data level of abstraction serve very usefully in many applications.
Regarding road vehicles, many advanced driver assistance systems (ADAS) are
possible thanks to multi-sensor data fusion filters. However, many applications
require higher level of abstraction to describe situations in which a vehicle is in-
volved. For military purposes, the issue of vehicle situation awareness has been
approached by several authors from the point of view of the artificial intelligence
[10]-[14]. Most of these authors agree to divide multi-sensor data fusion into four
levels of increasing situation complexity. Some other approaches, like the one
proposed by University of Melbourne, prefer different architecture schemes not
necessarily oriented to military scenes [15]. Other authors such as [16] proposed
a fusion architecture based on three levels of perception and functional areas,
like the one shown in Fig. 4. In this model, vertical axis represents perception