Page 439 - Automotive Engineering Powertrain Chassis System and Vehicle Body

CHAP TER 1 4. 2 Decisional architecture

Fig. 14.2-9 HILARE architecture (LAAS).

be able to efficiently and safely control the motions of a similar sequence of actions (i.e. which mainly differ
a car moving on the road network; another constraint is from one to the other by the execution parameters) each
to obtain smooth motions, i.e. to appropriately control time that the vehicle is going to execute an overtaking or
velocities and accelerations. a parallel parking manoeuvre. This approach is described
In the Sharp architecture, the three previous func- in more detail in Section 14.2.3.
tional layers are respectively represented by the Planner,
the Mission Scheduler and the Motion Controller (see 14.2.2.5 Conclusion
Section 14.2.3.1 for more details). However, the effi-
ciency and robustness of the system has been improved Nowadays, the three-layer paradigm represents a con-
by introducing a new concept for constructing on-line the sensus about the conception of a hybrid control archi-
action plans: the concept of ‘sensor-based manoeuvre’ tecture. Yet, the level of abstraction, of reactivity, and of
(SBM) which can be seen as meta-skill. Using this capacity of these layers are still a design problem and
approach, the Mission Scheduler can efficiently construct a motivation for many research works. This is probably
motion plans (i.e. with the required response time) by due to the fact that the nature of the integration of de-
combining previously planned trajectories with appro- liberative and reactive features is not yet well
priate instantiations of these generic skills (SBM). This understood; it is also probably due to the fact that no
approach has been both motivated and justified by the universal solution exists and that a particular type of con-
fact that in the considered application (i.e. an intelligent trol architecture responds to a particular niche of appli-
vehicle moving on the road network), the number of cation. The control architecture presented in Section
different types of manoeuvres to execute is finite and not 14.2.3 has specially been designed and implemented for
too large. In particular, it is useless to fully re-plan the intelligent vehicle application.

446