Information Technology Reference
In-Depth Information
the strict framework of a
vertical cooperation
(the driver is the only master on board
and the automaton only intervenes under his control). In order to subscribe to the
logic of
horizontal cooperation
[MIL 95], aiming to give the machine real decision-
making powers that are potentially very useful, particularly in a critical situation.
These questions, which reveal the ergonomics as much as the cognitive
engineering [BOY 03], are at the heart of current discussion in terms of driving
assistance. They were partially addressed by the authors of this chapter via different
research thematics that we present in the following section. The first thematic deals
with the prevention of collision risks and seeks to differentiate situations in the
context of adequately managing the risk, to those where assistance would be
required. The second concerns making the trajectory secure in its lateral dimension,
and in particular enabling us to explore the different modalities of human-machine
cooperation in depth. The last one addresses the question of the diagnosis of certain
human failures, such as inattention or falling asleep, in real time.
5.3. Three realizations of cooperative devices for the purposes of automobile
copiloting
5.3.1.
Preventing collision risks
The avoidance of collisions is a major stake for road safety. If we only consider
the risks of collisions between vehicles, the figures for the national observatory of
road safety
5
indicate that in France this type of collision represented more than 63%
of corporal accidents in 2007 (that is 51,548 cases out of a total of 81,272 accidents)
and more than 51% of fatal accidents on the road (2,385 people killed out of 4,620).
If we integrate collisions with pedestrians, these numbers go to nearly 80% for
corporal accidents (64,526) and 63% for fatal accidents (2,911), respectively.
In terms of human-machine cooperation, collision avoidance is a particularly
delicate question. It is a matter of appreciating the critical nature of the road
situation, which is in itself a complex task from a technological point of view
(extract targets relevant to the road scene and estimate the probability of occurrence
and imminence of trajectory conflicts). Nonetheless, this is far from being sufficient.
If we exclude the extreme conditions for which the time to collision is so brief that
no human intervention is likely to avoid the accident, all the configurations require
the activity of the driver to be taken into account in view of appreciating the
effective risk of collision with precision.
5 ONSIR: http://www2.securiteroutiere.gouv.fr/infos-ref/observatoire/accidentologie/le-bilan-
de-l.html.
