Toward a Generic Model for Autonomy Levels for Unmanned Systems (ALFUS) Hui-Min Huang, Elena Messina, James Albus
National Institute of Standards and Technology
Intelligent Systems Division
泰戈尔诗选Gaithersburg, Maryland 20899-8230
{ hui-min.huang, ssina, james.albus }@v
Abstract
Unmanned systems (UMSs) have been deployed to military and civilian operations. UMSs vary widely in their capabilities and purpos. It would be beneficial to have a t of widely recognized standard definitions on the capabilities of the UMSs. Efforts have begun in various organizations in defining autonomy levels for unmanned systems.
As part of this ongoing rearch, we are attempting to define a generic model for the autonomy levels for unmanned systems (ALFUS). Our intention is for this model to be ud to derive mission-s
pecific ALFUS. In this paper, we describe autonomy levels in three tiers, subsystem, system, and system of systems (SoS). Within each tier, the levels of autonomy are further divided with the factors of task complexity, environmental complexity, human involvement, and a t of quality factors. The work prented is a snapshot of an ongoing process.
1.Introduction
The technological advances in mobile robotics have been significant enough to warrant the deployment of unmanned systems (UMSs) in military and civilian operations. Aerial and undera UMSs have been performing missions for a number of years. Ground UMSs have been introduced in recent wars. UMSs have also participated in the arch and rescue missions after natural disasters and terrorist attacks.
UMSs vary widely in their capabilities and purpos. They may be developed either with particular requirements or for general purpo. Some operating environments are known and structured, allowing UMSs to perform repetitive but unsupervid tasks. Other environments are much more unpredictable, requiring unmanned systems to make decisions according to the current environmental conditions obtained through onboard nsing and processing capabilities. There is n
erar怎么用o guidance available to match situations with required system capabilities. Organizations planning to fund development of new autonomous systems currently lack means of specifying the level of autonomy required – and of validating that the delivered systems meet tho specifications.
It would, therefore, be beneficial to have a t of widely recognized standard definitions on the capabilities of the UMSs. ALFUS may address many aspects of the UMSs problem area, including as references for the system specification and performance measurement purpos. A particular procurement program may specify ALFUS-5. A particular area in a map may be certified for ALFUS-3 operations. Similar practices are en in other fields, such as the five-level Capability Maturity Model (CMM) in Software Engineering (although that applies to the organization that develops the software) [4] and the nine-level NASA/Army Technology Readiness Level (TRL) structure [5].
In this rearch, we attempt to develop a generic framework for the autonomy level specification for unmanned systems. Our objective is for particular system urs to be able to generate their specific ALFUS model from this generic model.
2.Related Work
There have been other discussions on autonomy levels published, but to our knowledge, there has b
een no other concerted effort to bring together communities of urs to define a t of autonomy measures that are common to the UMS constituency.
A definition of autonomy propod by Antsaklis et al. [6] states
Autonomous control systems must
perform well under significant
uncertainties in the plant and the
environment for extended periods of
time and they must be able to
compensate for system failures
without external intervention.
Ziegler describes conceptual views of autonomy from the perspective of veral fields (artificial intelligence, intelligent control, simulation, robotics, etc.) and propos a summary 3 level categoriza
tion [7]:
1.ability to achieve prescribed objectives, all
knowledge being in the form of models, as
in the model-bad architecture.
2.ability to adapt to major environmental
changes. This requires knowledge enabling
the system to perform structure
reconfiguration, i.e., it needs knowledge of
structural and behavioral alternatives as can规格英语
be reprented in the system entity
structure.
3.ability to develop its own objectives. This
requires knowledge to create new models to
support the new objectives, that is a
modeling methodology.
Whereas this broad classification is uful as a high-level abstraction of the categorization of capabilities, it would not provide much guidance to an Army procurement specification.
A more fully developed framework for defining Autonomous Control Levels (ACL) for air vehicles has been developed by the Air Force Rearch Laboratory. Clough [1] describes an 11 level ACL chart that ranges from 0 autonomy for remotely piloted vehicles to 10 for Human-Like. The highest level attainable by aerial vehicles is 9 for Multi-Vehicle Tactical Performance Optimization. There are various dimensions considered in determining the autonomy level: Perception/Situation Awareness, Analysis/Decision Making, and Communication/Cooperation. The model is specific to air vehicles. The Army Science Board has conducted a study on Human Robot Interface (HRI) [8]. This study includes an autonomy level chart covering a lowest level as remote control to a highest level as auto
nomous conglomerate. While this chart provides a good reference for generic levels, it lacks rationale and in-depth description guiding specific system urs. This has a potential of resulting in mis-identification of levels within particular programs.
The U.S. Army Maneuver Support Center and the National Institute of Standards and Technology (NIST) have earlier efforts of autonomy level charts [2, 3] that were developed for the Army Future Combat System (FCS) program. The charts provided references as how an end product of the autonomy level model may look like as we start our rearch effort from a generic point of view.
3.Defining Autonomy
The term autonomy must be defined before the autonomy level definitions. Section 2 described related definitions for certain contexts. Our definition for autonomy would be specific for unmanned systems.
3.1 A propod definition
We propo to define a UMS’s autonomy as its own capability to achieve its mission goals. The more complex the goals are, the higher the level of autonomy the UMS has achieved.
3.2UMS capability
We propo to define that levels of autonomy for a UMS are proportional to the system’s capability to perceive, plan, decide, and act to achieve the goals. Humans can play different types of roles [9, 10], which should affect the system autonomy in different ways. This issue will be investigated, in detail, in the next version of this model. In this version, we simplify the human interaction issue by stating that the amount and the criticality of human interactions would be inverly proportional to the levels of autonomy for a UMS. In other words, systems with high levels of autonomy, in general, require less human interactions. Further, the required human interactions should be non-critical and less in amount. The system with high levels of
autonomy should not only be able to perform routine missions independently, but also gracefully handle unexpected situations.
We make a distinction between the terms of “degrees of autonomy” and “levels of autonomy.” Total autonomy in low-level creatures does not correspond to high levels of autonomy. Examples include the movements of earthworms and bacteria that are 100% autonomous but considered low levels of autonomy.
3.3Goals and goal reprentations
Humans evaluate a system’s autonomy. Therefore, UMS stakeholders specify the system mission or task goals, which, in turn, dictate the system’s autonomy level. For a military Unmanned Ground Vehicle (UGV), the main areas of concern include mobility and mission behaviors. The mobility goals could be any tactical movement. Mission behaviors could include countermining. A UMS could have a high level of autonomy in mobility but a rudimentary level in any mission behaviors.
Within the context of mobility autonomy level, fuel sufficiency should be assumed. At a higher level of SoS autonomy, there should be a task of fuel management as a part of the mission activities.
The system’s goals, state, and status must be described and prented in reference frames that are easily understandable to humans. In an electro-mechanical system, the low-level actuator motions typically reference local coordinate frames. Once the components are asmbled into a subsystem, the subsystem-centric reference is typically ud. The same principle applies when subsystems are asmbled into a system and when a group of systems is asmbled into a SoS.
合肥日语
3.4Complexity
We propo to characterize the complexity of a UMS’s actions with the following attributes:
•Degrees of dynamics of the system actions.
When a UMS has to navigate through
complex trajectories with accelerations,
decelerations, turns, and stops, the actions
are considered more complex than
navigating through a smooth and straight
path without encountering any objects.
•Degrees of environmental uncertainly—frequencies of changes, visibility of objects.
•Number of steps toward the solutions.
•Relative amount of efforts for the UMS to decide and act—when a UMS is able to
respond to situation A quickly but to
situation B slowly, situation B might be
more complex with respect to the UMS’s
capability.
•Number of components involved in coordination.
•Levels of coordination—frequency of interactions.
4.Defining ALFUS国际部高中是什么
4.1Structure
We propo a multiple-layer generalization-specification structure for ALFUS. The objective would be the ability to instantiate the generic model for any mission or goals specific ALFUS as particular programs require.
Figure 1: Autonomy level model relationships Quite a few of the current solutions u a ten level structure; we prefer to relax this constraint to give us more freedom.
4.2Performance and quality factors
In trying to define various levels of autonomy, there are associated variables of:
•Mission behavior resolutions—for example, contrast a “road march from city
to city” vs. “go 50 meters on the road (with
traffic)”
•Perception capability—what level of perception capability do we assume for the
UMS to be able to perform specified
behaviors? For example, if we specify that a
UGV is to perform road following at
ALFUS-n, what kind of road conditions do
we imply? Is it sufficient to quantify the
conditions as excellent, fair, and poor?
Urs can further define the t of
quantification indices.
•Spatial and temporal resolutions—the example of pure road following may be
easy, i.e., all legal speeds, adding traffic
negotiation will be hard to specify. If we
define what constitutes an obstacle for an
ALFUS level, how far ahead is the obstacle
before the system detects it? How fast is
the UMS moving?
•Tolerance on the specified behavior—in performing a countermine operation, is a
夫人英语UMS categorized for a specified level if it
clears 85% of the field on average?
•Mission success rate—if a UMS accomplishes its missions 50% of the time,
how do we specify its autonomy level?
•Action quality and value judgment—does
a UMS that performs missions with optimal
solutions have a higher level of autonomy
than a UMS that completes missions with
only adequate solutions? Can the system
generate actions bad on risk and benefit
factors?
4.3Propod generic ALFUS
We describe our propod model for specifying UMS autonomy levels:
A.We propo that system autonomy should be
specified in four tiers: actuator, subsystem,
system, and System of Systems (SoS). The
actuator tier reprents remote control, the lowest level of autonomy. At the subsystem
tier, autonomy levels are defined in terms of
subsystem functions, such as mobility and
functions performed by onboard mission packages. The vehicle can be a part of an SoS
and the SoS, itlf, can have its own level of
autonomy. SoS can be further subdivided to
align with operational units. Error! Reference
source not found. depicts the structure with the
actuator tier omitted for simplicity.
B.Within each tier, autonomy levels are divided,
from low to high, according to following types
of actions:
•Simple goal attainment—when the UMS can only perform feedback actions with
respect to the UMS state (position in the
situation of mobility) to reach the given
goal and is unable to respond to any
environmental conditions1.
•When an individual UMS is able to achieve goals in a static environment. Examples in
the UGV domain would include road
following and obstacle avoidance. The
UMS posss required perception
capability to perform the behaviors.
•When a group of UMSs is able to achieve goals in a static environment.
•When an individual UMS is able to achieve goal in a dynamic environment. An
example in the UGV domain would be on-
road driving in traffic.
•When a group of UMSs is able to achieve goal while in a dynamic environment. An
example would be when a military unit of
UMS is fighting with an enemy unit, both
in motion.
Each higher level assumes all the capabilities stated for the low levels. Figure 3 illustrates the structure.
C.Each level, as specified in paragraph B, can be
further subdivided according to the performance
and quality factors as stated in ction 4.2.
1 For SoS, this would be to move a SoS to location
without concerning group level obstacles such as enemy
units. Figure 2: ALFUS Classification
reliableFigure 3: Dividing ALFUS per environmental difficulties
There could be multiple ways of characterizing
the factors. For example, mission success rate can be reprented in terms of either percentage, a high-resolution way, or from one
to ten, a low resolution way. Since the purpo
of this rearch is to quantify system autonomy
in a low-resolution scale, typically from one to
ten, all the performance and quality factors should be characterized in low-resolution indices. The values for each factor can be either
individually kept or lumped into a single index,
depending on urs requirements. Figure 4 illustrates that the factors are combined to divide system autonomy levels.
Figure 4: Further dividing subsystem ALFUS per quality factors
Human factors can divide the autonomy levels
similarly. Further elaborations will be conducted in the next version.
Figure 5: ALFUS to be mission specific
D.Our propod ALFUS model is bad on
missions and tasks. Urs specify mission requirements, which should be analyzed and form task structures according to complexity.hgf
大森美玲
In military domains, the task structures should be parallel to or consistent with commanding structures and are ud as a basis for specifying
system autonomy levels. Figure 5 shows that, low level missions are specified for subsystems.
They are integrated into high-level missions as the system scope expands.
4.4Example
Figure 6 shows a simplified view of possible autonomy levels bad on task complexity. The mission is to make sure that the NIST campus is cure. At each level, a possible task is shown for that level, along with a descriptive list of the attributes for the capabilities, knowledge levels, uncertainty assumptions, and spatial and temporal scopes. The attributes illustrate the complexity of the tasks. The type of interaction (level of discour) becomes more dependent on human decision-making and intelligence in the lower levels of the hierarchy.
5.Summary
We define a preliminary model for specifying levels of autonomy for unmanned systems. The model contains a definition of autonomy, a t of tiers for autonomy per system configuration, and a t of v
ariables for dividing autonomy levels. An example is ud to illustrate some aspects of this model. We plan to further develop this model and
verify the model against particular applications.
Note that all the high, medium, low, large, and small designations are relative.
Figure 6: Example of levels of autonomy characteristics for an unmanned system
References
1.Bruce T. Clough , “Metrics, Schmetrics! How The
Heck Do You Determine A UAV’s Autonomy Anyway?” Proceedings of the Performance Metrics
for Intelligent Systems Workshop, Gaithersburg, Maryland, 2002.
2.Knichel, David, Position Prentation for the
Maneuver Support Center, Directorate of Combat Development, U.S. Army, the First Federal Agencies
Ad Hoc Working Group Meeting for the Definition
of the Autonomy Levels for Unmanned Systems, Gaithersburg, MD, July, 18, 2003.
3.James Albus, Position Prentation for National
Institute of Standards and Technology, Intelligent Systems Division, the First Federal Agencies Ad Hoc Working Group Meeting for the Definition of
the Autonomy Levels for Unmanned Systems, Gaithersburg, MD, July, 18, 2003.
4.u.edu/cmm/cmm.html
5.John C. Mankins, Technology Readiness Levels: A
White Paper. Office of Space Access and Technology,NASA,
www.v/office/codeq/trl/trl.pdf,
April 6, 1995.kululu
6.Antsaklis, P.J., K.M. Passino and S.J. Wang,
“Towards Intelligent Autonomous control Systems: Architecture and Fundamental Issues,” Journal of Intelligent and Robotic Systems, Vol. 1, No. 4, pp.
315-342.
7.Zeigler, B.P., “High autonomy systems: concepts
and models,” AI, Simulation, and Planning in High Autonomy Systems, 1990, Proceedings, 26-27 Mar 1990.
8.Army Science Board, Ad Hoc Study on Human
Robot Interface Issues, Arlington, Virginia, 2002.
9.Parasuraman, R., Sheridan, T. B., and Wickens, C.
D. A Model for Types and Levels of Human
Interaction with Automation. IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans, Vol. 30, No. 3 May 2000.
10.Drury, J., Scholtz, J., and Yanco, H., “Awareness in
Human-Robot Interactions,” to appear in the Proceedings of the 2003 IEEE Systems, Man, and Cybernetics conference, Washington DC, October 2003.
