Information Technology Reference
In-Depth Information
independently of each other (particularly if these pieces of equipment are integrated
by car manufacturers themselves). This “cumulative” logic of assistance, designed
in juxtaposition to autonomous systems, is quickly reaching its limits. At the end of
the day , indeed, there is only one driver whose attention must primarily be on the
road, and for whom the integration of assistance is a more than imperious
requirement. Conversely, there is not only a major risk of perceptive and cognitive
overload, but it can also be a source of human error (ambiguity regarding the origin
or effective sense of certain information, potential confusion from the point of view
of commands to be activated, etc.). This problem, though it has already been
identified in the field of embedded information systems, also deserves to be
investigated in depth regarding devices for the automation of driving.
What would happen if two different devices recommended or engaged
contradictory driving actions at the same time? (This could include breaking to
respect a speed limit versus accelerating to safely join another stream of traffic,
keeping the vehicle in the lane to guarantee a safe trajectory versus leaving the lane
in order to avoid an obstacle, etc.) Ideally, it would be advisable to design these
assistants in synergy with each other from the start: a cruise control device would
have a lot to gain from being part of a collision avoidance function, for example
when a monitoring situation suddenly becomes critical. This would be all the more
relevant as these two assistance functions, artificially disassociated from a
technological point of view, are in continuum from the point of view of human
activity.
This is not necessarily easy to implement from an industrial point of view, and
the “integrated” and “organic” development of all the assistance functions in a
single device is no doubt not realistic, in the short term anyway. However, there is
nothing to say that it is not possible to move forward in a more pragmatic fashion,
based on the logic of distributed intelligence (to which we will return in the
conclusion). This is headed by the first-rate assistance devices of copiloting meta-
function, designed as an integrative layer in charge of guaranteeing efficient
coordination of driving aids in favor of harmonious cooperation with the human
driver.
This integrative dimension of an automobile copilot must furthermore be
completed by an adaptive dimension . As we specified in the introduction, not all
drivers need to be assisted in the same way or at the same time. The machine must
therefore “listen” to the human operator, all the while taking into account the
requirements of the driving situation. Ultimately, it is therefore a matter of heading
towards ergonomics “in real time” via the adaptive nature of assistants, to provide a
contextualized aid conscious of the effective needs and the potential difficulties that
the driver might have at a moment in time t . An important stake at this level resides
in the ability of the copilot to adopt a critical eye regarding human decisions and
Search WWH ::




Custom Search