C h a p t e r 5
Looking to the future, it's clear that the
concept of Physical AI will shape the next
evolution of AMRs.
Physical AI refers to using AI to directly
interact with and understand the
physical world, encompassing everything
from sensing and understanding the
environment to manipulating objects
and making decisions based on real-time
data. These robots will not be confined to
predefined tasks or static environments.
Instead, they will learn from context,
adjust to unstructured surroundings,
and generalize their capabilities across
a wide range of use cases. Importantly,
such a shift is underpinned by a transition
from rule-based autonomy to behavior-
driven intelligence. This is powered
by foundation models and real-time
environmental feedback.
On a high level, physical AI combines
perception, planning, and actuation
into a unified feedback loop. This
convergence allows robots to understand
both spatial and semantic elements
of their environment. For example, a
robot that recognizes a table and drives
around it must also understand that
an object falling off the table has not
disappeared, but is rather lying on the
ground. Any object that is out of the
perception of the robot is still somewhere
THE FUTURE OF AMRS AND
PHYSICAL AI
The real significance of embodied
AI and foundation models is their
potential to eliminate the constant
trade-offs between flexibility and
reliability in AMR deployments.
Instead of over-engineering for edge
cases or limiting functionality, we
can now train models that adapt and
scale with the operation."
Victoria Quinde
System Engineering Manager, Dematic
23
Engineering the Future: The Sensors and Systems Powering Modern Mobile Robots
