Autonomous and
semiautonomous systems are characterized by their decision-making dependence on
external orders.
They have something in common with conditionally autonomous
and automated systems.
Semiautonomous systems depend on a human user somewhere
"in the loop" for decision-making, behavior management, or contextual
interventions, while autonomous systems may make decisions within a defined
region of operation without human input.
Under some situations, conditionally autonomous systems may
operate independently.
Automated systems differ from semiautonomous and autonomous
systems (autonomy) (automation).
The actions of the earlier systems are preset sequences
directly related to specific inputs, while the later systems' actions are
predefined sequences directly tied to specified inputs.
When a system's actions and possibilities for action are
established in advance as reactions to certain inputs, it is termed automated.
A garage door that automatically stops closing when a sensor
detects an impediment in its path is an example of an automated system.
Sensors and user interaction may both be used to collect
data.
An automated dishwasher or clothes washer, for example, is a
user-initiated automatic system in which the human user sets the sequences of
events and behaviors via a user interface, and the machine subsequently
executes the commands according to established mechanical sequences.
Autonomous systems, on the other hand, are ones in which the
capacity to evaluate conditions and choose actions is intrinsic to the system.
The autonomous system, like an automated system, depends on
sensors, cameras, or human input to give data, but its responses are marked by
more complicated decision-making based on the contextual evaluation of many
simultaneous inputs such as user intent, environment, and capabilities.
When it comes to real-world instances of systems, the terms
automated, semiautonomous, and autonomous are used interchangeably depending on
the nature of the tasks at hand and the intricacies of decision-making.
These categories aren't usually defined clearly or exactly.
Finally, the degree to which these categories apply is
determined by the size and scope of the activity in question.
While the above-mentioned basic differences between automated,
semiautonomous, and autonomous systems are widely accepted, there is some
dispute as to whether these system types exist in real systems.
The degrees of autonomy established by SAE (previously the
Society of Automotive Engineers) for autonomous automobiles are one example of
such ambiguity.
Depending on road or weather conditions, as well as
situational indices like the existence of road barriers, lane markings,
geo-fencing, adjacent cars, or speed, a single system may be Level 2 partly
autonomous, Level 3 conditionally autonomous, or Level 4 autonomous.
The degree of autonomy may also be determined by how an
automobile job is characterized.
In this sense, a system's categorization is determined as
much by its technical structure as by the conditions of its operation or the
characteristics of the activity focus.
EXAMPLES OF AUTONOMOUS AI SYSTEMS
E Vehicles that are self-driving.
The contrasts between
automated, semiautonomous, conditionally autonomous, and completely autonomous
vehicle systems are shown using automated, semiautonomous, conditionally
autonomous, and fully autonomous car systems.
Automated technology, like as cruise control, is an example.
The user specifies a vehicle speed goal, and the vehicle
maintains that speed while adjusting acceleration and deceleration as needed by
the terrain.
However, in the case of semiautonomous vehicles, a vehicle
may be equipped with an adaptive cruise control feature (one that regulates a
vehicle's speed in relation to a leading vehicle and to a user's input), as
well as lane keeping assistance, automatic braking, and collision mitigation
technology.
Semiautonomous cars are now available on the market.
Many possible inputs (surrounding cars, lane markings, human
input, impediments, speed restrictions, etc.) may be interpreted by systems,
which can then regulate longitudinal and latitudinal control to
semiautonomously direct the vehicle's trajectory.
The human user is still involved in decision-making,
monitoring, and interventions in this system.
Conditional autonomy refers to a system that allows a human
user to "leave the loop" of control and decision-making under certain
situations.
The vehicle analyzes emergent inputs and controls its
behavior to accomplish the objective without human supervision or intervention
after a goal is set (e.g., to continue on a route).
Internal to the activity (defined by the purpose and
accessible methods), behaviors are governed and controlled without the
involvement of the human user.
It's crucial to remember that every categorization is
conditional on the aim and activity being operationalized.
Finally, an autonomous system has fewer constraints than
conditional autonomy and is capable of controlling all tasks in a given
activity.
An autonomous system, like conditional autonomy, functions
inside the activity structure without the involvement of a human user.
Autonomous Robotics
For a number of reasons, autonomous
systems may be found in the area of robotics.
There are a variety of reasons why autonomous robots should
be used to replace or augment humans, including safety (for example,
spaceflight or planetary surface exploration), undesirable circumstances
(monotonous tasks such as domestic chores and strenuous labor such as heavy
lifting), and situations where human action is limited or impossible (search
and rescue in confined conditions).
Robotics applications, like automobile applications, may be
deemed autonomous within the confines of a carefully defined domain or activity
area, such as a factory assembly line or a residence.
The degree of autonomy, like autonomous cars, is dependent
on the specific area and, in many situations, excludes maintenance and repair.
Unlike automated systems, however, an autonomous robot
inside such a defined activity structure would behave to achieve a set
objective by sensing its surroundings, analyzing contextual inputs, and
regulating behavior appropriately without the need for human interaction.
Autonomous robots are now used in a wide range of
applications, including domestic uses such as autonomous lawn care robots and
interplanetary exploration applications such as the Mars rovers MER-A and
MER-B.
Semiautonomous Weapons
is an acronym for
"Semiautonomous Weapons." As part of contemporary military
capabilities, autonomous and semiautonomous weapon systems are now being
developed.
The definition of, and difference between, autonomous and
semiautonomous changes significantly depending on the operationalization of the
terminology, the context, and the sphere of activity, much as it does in the
preceding automobile and robotics instances.
Consider a landmine as an example of an automated weapon
that is not self-contained.
It reacts with fatal force when a sensor is activated, and
there is no decision-making capabilities or human interaction.
A semiautonomous system, on the other hand, processes inputs
and acts on them for a collection of tasks that form weaponry activity in
collaboration with a human user.
The weapons system and the human operator must work together
to complete a single task.
To put it another way, the human user is "in the
know." Identifying a target, aiming, and shooting are examples of these
activities.
Navigation toward a target, placement, and reloading are all
possible.
These duties are shared between the system and the human
user in a semiautonomous weapon system.
An autonomous system, on the other hand, would be
accountable for all of these duties without the need for human monitoring,
decision-making, or intervention after the objective was determined and the
parameters provided.
There are presently no completely autonomous weapons systems
that meet these requirements.
These meanings, as previously stated, are technologically,
socially, legally, and linguistically dependent.
The distinction between semiautonomous and autonomous
systems has ethical, moral, and political implications, particularly in the
case of weapons systems.
This is particularly relevant for assessing accountability,
since causal agency and decision-making may be distributed across developers
and consumers.
As in the case of machine learning algorithms, the sources
of agency and decision-making may also be ambiguous.
USER-INTERFACE CONSIDERATIONS.
The various obstacles
in building optimum user interfaces for semiautonomous and autonomous systems
are mirrored in the ambiguity of their definitions.
For example, in the case of automobiles, ensuring that the
user and the system (as designed by the system's designers) have a consistent
model of the capabilities being automated (as well as the intended distribution
and degree of control) is crucial for the safe transfer of control
responsibility.
In the sense that once an activity area is specified,
control and responsibility are binary, autonomous systems pose similar
user-interface issues (either the system or the human user is responsible).
The problem is reduced to defining the activity and
relinquishing control in this case.
Because the description of an activity domain has no
required relationship to the composition, structure, and interaction of
constituent activities, semiautonomous systems create more difficult issues for
the design of user interfaces.
Particular tasks (such as a car maintaining lateral position
in a lane) may be decided by an engineer's use of specific technical equipment
(and the restrictions that come with it) and therefore have no link to the
user's mental representation of that work.
An obstacle detection task, in which a semiautonomous system
moves about an environment by avoiding impediments, is an example.
The machine's obstacle detection technologies (camera,
radar, optical sensors, touch sensors, thermal sensors, mapping, and so on)
define what is and isn't an impediment, and such restrictions may be opaque to
the user.
As a consequence of the ambiguity, the system must
communicate with a human user when assistance is required, and the system (and
its designers) must recognize and anticipate any conflict between system and
user models.
Other considerations for designing semiautonomous and
autonomous systems (specifically in relation to the ethical and legal
dimensions complicated by the distribution of agency among developers and
users) include identification and authorization methods and protocols, in addition
to the issues raised above.
The difficulty of identifying and approving users for
autonomous technology activation is crucial since once activated, systems no
longer need continuous monitoring, intermittent decision-making, or
interaction.
~ Jai Krishna Ponnappan
You may also want to read more about Artificial Intelligence here.
