Page 27 - Dynamic Vision for Perception and Control of Motion

1.5 What Type of Vision System Is Most Adequate? 11

1.4.3 Global Integrals for Situation Assessment

More complex situations encompassing many objects or missions consisting of se-
quences of mission elements are represented in the lower right corner of Figure
1.2. Again, how to best choose the subdivisions and the absolute scales on the time
axis or in space depends very much on the problem area under study. This will be
completely different for a task in manufacturing of micro-systems compared to one
in space flight. The basic principle of subdividing the overall task, however, may
be according to the same scheme given in Figure 1.2, even though the technical
elements used may be completely different.
On a much larger timescale, the effect of entire feed-forward control time histo-
ries may be predicted which have the goal of achieving some special state changes
or transitions. For example, lane change of a road vehicle on a freeway, which may
take 2 to 10 seconds in total, may be described as a well-structured sequence of
control outputs resulting in a certain trajectory of the vehicle. At the end of the ma-
neuver, the vehicle should be in the neighboring lane with the same state variables
otherwise (velocity, lateral position in the lane, heading). The symbol “lane
change”, thus, stands for a relatively complex maneuver element which may be
triggered from the higher levels on demand by just using this symbol (maybe to-
gether with some parameters specifying the maneuver time and, thereby, the
maximal lateral acceleration to be encountered). Details are discussed in Section
3.4.
These “maneuver elements”, defined properly, allow us to decompose complex
maneuvers into stereotypical elements which may be pieced together according to
the actual needs; large sections of these missions may be performed by exploiting
feedback control, such as lane following and distance keeping for road vehicles.
Thereby, scales of distances for entire missions depend on the process to be con-
trolled; these will be completely different for “autonomously guided vehicles”
(AGVs) on the factory floor (hundreds of meters) compared to road vehicles (tens
of km) or even aircraft (hundreds or thousands of km).
The design of the vision system should be selected depending on the task at
hand (see next section).

1.5 What Type of Vision System Is Most Adequate?

For motion control, due to inertia of a body, the actual velocity vector determines
where to look to avoid collisions with other objects. Since lateral control may be
applied to some extent and since other objects and subjects may have a velocity
vector of their own, the viewing range should be sufficiently large for detecting all
possible collision courses with other objects. Therefore, the simultaneous field of
view is most critical nearby.
On the other hand, if driving at high speed is required, the look-ahead range
should be sufficiently large for reliably detecting objects at distances which allow
safe braking. At a speed of 30 m/s (108 km/h or about 65 mph), the distance for
2
braking [with a deceleration level of 0.4 Earth gravity g (9.81 m/s , that is a x §í 4