Page 308 - Introduction to Autonomous Mobile Robots

Planning and Navigation

off-line planning 293

strategic decisions

tactical decisions

quasi real-time

hard real-time

Figure 6.18
Generic temporal decomposition of a navigation architecture.

affects the robot’s immediate actions and is therefore subject to some temporal constraints,
while a strategic or off-line layer represents decisions that affect the robot’s behavior over
the long term, with few temporal constraints on the module’s response time.
Four important, interrelated trends correlate with temporal decomposition. These are not
set in stone; there are exceptions. Nevertheless, these general properties of temporal
decompositions are enlightening:

Sensor response time. A particular module’s sensor response time can be defined as the
amount of time between acquisition of a sensor-based event and a corresponding change in
the output of the module. As one moves up the stack in figure 6.18 the sensor response time
tends to increase. For the lowest-level modules, the sensor response time is often limited
only by the raw processor and sensor speeds. At the highest-level modules, sensor response
can be limited by slow and deliberate decision-making processes.

Temporal depth. Temporal depth is a useful concept applying to the temporal window
that affects the module’s output, both backward and forward in time. Temporal horizon
describes the amount of look ahead used by the module during the process of choosing an
output. Temporal memory describes the historical time span of sensor input that is used by
the module to determine the next output. Lowest-level modules tend to have very little tem-
poral depth in both directions, whereas the deliberative processes of highest-level modules
make use of a large temporal memory and consider actions based on their long-term con-
sequences, making note of large temporal horizons.