=Paper=
{{Paper
|id=Vol-1523/STIDS_2015_T11_Mumford_etal
|storemode=property
|title=Toward Representing and Recognizing Cyber-Physical Autonomous Agents in Competition Using Event Semantics
|pdfUrl=https://ceur-ws.org/Vol-1523/STIDS_2015_T11_Mumford_etal.pdf
|volume=Vol-1523
|dblpUrl=https://dblp.org/rec/conf/stids/MumfordWC15
}}
==Toward Representing and Recognizing Cyber-Physical Autonomous Agents in Competition Using Event Semantics==
https://ceur-ws.org/Vol-1523/STIDS_2015_T11_Mumford_etal.pdf
Toward Representing and Recognizing
Cyber-Physical Elements in Competition Using
Event Semantics
Alonza Mumford, Duminda Wijesekera, Paulo Costa
George Mason University
amumford@gmu.edu, dwijesek@gmu.edu, pcosta@gmu.edu
Abstract—The Federal Aviation Administration (FAA) is ob- some real-world applications where the activities of human and
serving an increasing number of incidents involving recreational physical autonomous agents are identified by sensors coupled
drones, and imagining a future where every drone will be to a cyber-physical system:
equipped with an Automatic Dependent Surveillance-Broadcast
(ADS-B) transponder that communicates and cooperates with • In the National Football League (NFL), each football
the FAA’s Next Generation (NextGen) Aviation Cyber-Physical player and stadium is equipped with RFID sensors and
System in order to help mitigate aerial collision risk [1]. This receivers permitting the league to track fine-grained loca-
exemplar application involves human or autonomous agents tion data for each play. In this case, the Internet-of-Things
interacting within some sort of cyber-physical system where
competition or cooperation between cyber-physical elements (IoT) and CPS has been incorporated into operations and
exist. We anticipate that the use of higher-level abstractions management of professional sports venues [2].
will be required for modeling human or autonomous agent’s • In the recreational Unmanned Aerial Vehicle (UAV) mar-
interactions within these type of systems in order to make sense ket, some manufacturers have equipped their drones with
of the observations derived from sensor-data. In this paper, we ADS-B sensors, which is a type of sensor for aerial coop-
articulate an approach that uses event semantics to represent
the temporal, spatial, factor, and outcome features of activities erative collision detection and avoidance [3]. ADS-B is an
generated by competing or cooperating agents functioning within element of the Federal Aviation Administration (FAA)’s
a cyber-physical environment. We use those semantics, along with Next Generation Air Transportation System (NextGen),
observations of activity, to model higher-level activity abstractions which has been described as a airborne network instanti-
and to help perform strategy recognition from a concrete, ation of a cyber-physical system [1].
competition-oriented scenario reflected in a real-world, game data
set comprised of more than a half million events involving nearly These examples also illustrate a specific characteristic
8500 unique agents. The strength of the approach is grounded in present in some cyber-physical systems where competition
a specification of event semantics for our concrete multi-agent, or cooperation exist between human or physical autonomous
competitive game ontology using Resource Description Frame- agents. Subsequently, this paper reflects an interest in these
work Schema (RDFS) and Ontology Web Language (OWL). By
leveraging these Semantic Web languages, we anticipate that the
type of cyber-physical systems. This research effort is nar-
use of event semantics to describe cooperative or competitive rowly focused on identifying higher-level abstractions such
agent interactions within cyber-physical systems will become as strategies used by agents from observations derived from
more predominant in the future. sensors in the CPS. Further, we focus on an approach for
Index Terms—Agent-Based Model, Human Agent, Au- the representation and modeling of competitive actions and
tonomous Agent, Unmanned Aerial Vehicle, Gridiron Football,
interactions of agents in cyber-physical systems.
Semantic Web
II. M ETHODOLOGY
I. I NTRODUCTION The high-level methodology for this research activity has
Cyber-physical systems (CPS) are at the outset of com- been decomposed into five components. First, an exemplar for
pletely changing how society interacts with the physical world our experiment is identified. Gridiron Canadian and American
around it. These systems measure different features across Football is distinguished as an elaborate, competitive game
the physical environment (e.g., the location of an agent) and activity that involves multiple agents, which are organized
enable computational models that interact with a cyber-core into two teams for the purpose of executing a series of
(i.e., the computing and communications backbone of the CPS) offensive/defensive advances intended to score points and win
and with their corresponding physical environment to provide the game. This scenario is selected to match the CPS exemplar
some desired benefit or utility. In most cases, sensors provide identified in [2]. A NFL play-by-play data set that offers
the cyber-core with the primary mechanism for recognizing a likeness or model of the kind of RFID-derived data we
events or changes in the physical environment. The actions and would expect in [2] is acquired for the experiment. Second,
interactions between human or physical autonomous agents the information within the data set is conceptualized based on
and the cyber-core are captured through sensors. Consider the domain knowledge of Gridiron Football and modeled for
STIDS 2015 Proceedings Page 82
its semantic relationships using event semantic abstractions. ball, kickoffs, free kicks, scoring and penalties associated
The result is the contributed Gridiron Football Ontology, with Gridiron Football was acquired primarily through a
which is a conceptual vocabulary of American and Canadian literature review of popular publications such as Football’s
football, in the namespace http://www.ncsu.org/kr/fball. Third, Matchup Zone Coverages [4], The Art of Place-Kicking and
the data set is extracted from its previous form, transformed Punting [5], Defending the Spread Offense [6], Offensive
into the Resource Description Framework (RDF) metadata Football Strategies [7] and Winning Football [8]. Familiarity of
model (including serialization) according to the fball ontology these concepts was used to formulate ontological abstractions
and loaded into a public-accessible SPARQL Protocol and designed to codify types of events with associative kinds of
RDF Query Language (SPARQL) Endpoint and RDF Store. entity-attribute-models with agent, temporal, spatial, factor and
Fifth, we integrate or point a public-accessible SPARQL product classes and relationships in the Gridiron Football
Endpoint Explorer for Expressive Question Answering to the Ontology.
Gridiron Football Event Endpoint in order to interrogate the
nearly 40 million triples generated for indications of tactics or C. Formal Ontology Modeling
strategies being employed during game events. Further details The contributed Gridiron Football Ontology expresses a
are provided in each respective section of the paper. formal representation of knowledge within the Gridiron (or
American and Canadian) Football domain via a set of concepts
III. P ROPOSED S OLUTION and inter-concept relationships. The ontology is engineered
A. Data according to a methodology that partitions the ontology into
two levels: theory ontology (or upper ontology) and domain
Our research effort was presented with a considerable
ontology (or lower ontology). The theory ontology is abstract
challenge pertaining to the acquisition of rich context data
and compact. It focuses on concepts such as time, space, goals,
that would comprise human or agent activities, and that would
etc. The domain ontology provides a formal description of
simulate the type of data we would expect to acquire from the
the classes (i.e., concepts) and relationships between classes
scenario identified in [2].
that exist in a domain. In this manner, the Gridiron Football
Ontology is conceived as a two-level ontology with an upper
level that abstracts a football game as a sequence of orga-
nized spatio-temporal events and a lower level that provides
Fig. 1. An example illustration of play-by-play text generated by a NFL
statistician, which translates to: a RUSH play event occurred at the "7 minute a concrete specification of the ontology components (i.e.,
and 41 seconds" timestamp of the quarter; D. Williams rushed the ball in the individuals, classes, attributes, relations, etc.) associated with
direction of his Left Guard and progressed 6 yards up to the 13 yard line of Gridiron Football. By concrete specification, we mean that the
the opposing Pittsburgh Steelers team; and where he was jointly tackled by
A. Branch and G. Grissom. ontology is machine-readable and understandable. The result
is that RDF Schema (RDFS) Language and Web Ontology
Prior to recent introduction of RFID sensors into NFL game Language (OWL)-based ontological software components were
stadiums [2], the data capture of movements of players on a developed for all items (370+ unique key-value pairs) in
football game field demanded a human-in-the-loop to observe the NFL Play-by-Play data set. Ontological modeling and
the execution of plays across the field and to sequentially engineering efforts were assisted using several Semantic Web
report the specifics of each event as illustrated in figure 1. tools to include CMAP Knowledge Modeling Environment [9]
During games, this reporting data is manually generated by and Stanford Protege [10].
a distributed network of league statisticians, and immediately At the center of the upper ontology is the notion of an
propagated from individual stadiums to hundreds of websites event, and we reuse the existing Event Ontology developed
in a span of seconds [2]. In turn, the data set used in the at Queen Mary, University of London [11]. The ontology is
experiment is provided by a sports analysis firm that uses web partitioned into a set of classes and properties identified as:
extraction and other techniques to generate a structured data event:Event, which is an arbitrary classification of a space/time
set from disparate web post left by these league statisticians. region, which may have participating agents, passive factors,
The data set contains NFL play-by-play events from 2000 to products, and a location); event:sub_event, which provides a
2013. Further, we assert that this data is characteristic of the mechanism to partition a complex event; event:agent, which
type of complex yet coarse- or fine-grained event data that relates an event to an active agent such as a person; event:time,
can be expected from observations of human or autonomous which relates an event to a time object such as a duration of
physical agent’s activities as they engaged in competition time. event:place, which relates an event to a spatial object;
within a cyber-physical system. event:product, which relates an event to something produced
during an event; and event:factor, which relates an event to
B. Domain Knowledge Acquisition something that contributes to its result such as a cause.
An effort was made to grasp an understanding of the phys- In figure 2, domain-specific classes and properties are
ical environment in which our football player agents operate. created through subsumption of the event:Event class to
Domain knowledge of concepts such as game field, players, model the many type of events associated with Gridiron
game timeline, driving the ball, alternating possession of the Football. For example, fball:GameEvent class conceptualizes
STIDS 2015 Proceedings Page 83
Fig. 2. A detailed specification of the fball ontology’s Active Agent description using portions of the Event Ontology and Friend Of A Friend (FOAF)
vocabularies. In this context, an active agent is a person or machine that performs in an event. Ontological components that are members of the upper ontology
are illustrated in blue color whereas the members of the Gridiron Football domain ontology are colored in yellow.
STIDS 2015 Proceedings Page 84
Fig. 3. A detailed specification of the fball ontology’s active agent description using portions of the Event Ontology and Basic Geo vocabularies.
a football game whereas fball:SeasonEvent represents a Football. For example, fball:KickoffOutcome is related to
football season. In the case of the NFL, a season event fball:KickoffEvent using the fball:hasOutcome. Though not
is comprised of 16 football games over a duration of 17 shown, the event:Product and event:hasProduct are sub-
weeks. The fball:hasFootballSubEvent property is devised sumed to generate the various types of football outcomes
through subsumption of the event:hasSubEvent property and and fball:hasOutcome property respectively. In addition,
used to partition a Season event and a Game event into domain-specific classes and properties are devised by way
a collection of games and collection of plays respectively. of subsumption of the event:Factor class to model the
The foaf:Agent class and its sub-classes, foaf:Person many types of factors that affect different types of foot-
and foaf:Group are extended to model the various types ball events associated with Gridiron Football. For example,
of football playing positions (e.g., fball:LeftTackle), the fball:WeatherFactor and fball:FieldConditionFactor are re-
membership of particular playing positions to certain lated to fball:GameEvent using the fball:hasEventFactor. The
football sub-groups (e.g., a Left Tackle has membership event:Factor and event:hasFactor are subsumed to generate
of the fball:OffensiveLineUnit). The fball:Player is the various types of football factors and fball:hasEventFactor
established to represent the idea of a generic football property respectively.
player that has datatype properties (i.e., attributes) such as In addition, portions of the Time [OWL-TIME] [12] and
fball:hasFullName, fball:hasPrimaryPosition, fball:hasHeight, The WGS84 Geo Positioning Ontology [Basic Geo] [13]
fball:has40YardDashTime and fball:hasBenchPressWeight. vocabularies are used. The primary classes and properties
The fball:hasPlayer, fball:hasUnit and fball:hasTeam include: time:TemporalEntity, which is a parent class that
properties are created by means of subsumption of the relates temporal information to an event; time:Interval, which
event:hasAgent property to relate football persons and groups is a subclass of TemporalEntity and temporal things with
to kinds of football events. For example, the fball:hasPlayer extent that have interior points; and geo:SpatialThing, which
property is used to establish a relation between the is a parent class that relates spatial information to an event.
fball:FumbleEvent and one of three player roles expected For brevity, the time and spatial ontological components used
or required during a fumble type of event: fball:Fumbler, in the fball ontology are not described; however, spatial
fball:ForcingPlayer and fball:RecoveringPlayer. ontological components are illustrated in figure 3.
Domain-specific classes and properties are devised by D. Metadata (RDF) Creation
way of subsumption of the event:Product class to model This work involved creating efficient RDF representations
the many types of outcomes that result from differ- for the NFL Play-by-Play data set using the Gridiron Football
ent types of football events associated with Gridiron Ontology. The effort experimented with reasoning software
STIDS 2015 Proceedings Page 85
Fig. 4. An illustration of the SPARKLIS interface/engine being used to interrogate the RDF triples generated from the Gridiron Football Ontology and NFL
data set. In this example, the attributes of location, time, factor, outcome and agent entities associated with this particular NFL play event are explored.
for the purpose of providing an inference reasoning capability uses to counter his opponent’s defensive attack. Here, we show
to the project’s software components for deriving new RDF multiple components that may be decomposed from a team’s
triples (knowledge) from the instances generated directly from overall offensive scheme according to [17]:
input data and its related ontology. This activity involved
• Running Component: Man/Power Blocking, Zone Block-
software development using the Apache Jena Open Source
ing and Flex Blocking
Java Framework for Semantic Web Development [14] to covert
• Passing Component (Setup Mode): Run to Setup the Pass,
the elaborate NFL play-by-play data set, which is comprised of
Pass to Setup the Run and Take What the Defense Gives
more than a half million game events and nearly 8500 unique
You
agents, to a semantic graph containing 44,676,644 RDF triples.
• Passing Component (Tempo): Normal Tempo and Hurry-
E. Metadata Storage and Retrieval Up Tempo
• Passing Component (Huddle): Normal Huddle and No
Specifically, this effort consisted of the deployment a
Huddle
SPARQL End-Point web server and RDF-based triple store
• Passing Component (Length of Passes): Short to Interme-
using the Apache Fuseki/TDB suite [15]. Ingest of triples
diate and Vertical Intermediate to Deep Passing Game
into the triple store are primarily made by scripts or the
• Passing Component (Quarterback Position): Under Cen-
upload feature in the Apache Fuseki web client. Queries
ter and Pistol Depth, Shotgun Depth
are made through also made through the SPARQL interface
• Passing Component (Route Assignments): Route Tree
within the Fuseki client as well as the Sparklis: a SPARQL
Assignments (Air Coryell), Group Assignments (Erhardt-
Endpoint Explorer for Expressive Question Answering [16]
Perkins)
web service. In Figure 4, an illustration of SPARKLIS being
• Passing Component (How, Where, When): Predetermined
used to interrogate the generated NFL Football triples is given.
Pass to Spot Before Break, Predetermined Pass to a
IV. ACTIVITY AND S TRATEGY R ECOGNITION Person after the Break and Option Pass to a Person after
the Break
In this section, we apply our approach of ontology-based
activity recognition to the Gridiron Football domain and try In our experiment, we attempt to identify these components
to show how event semantics may be used to help identify the of an offensive scheme. This is at least partially achieved by
base offensive scheme being used by a particular team during integrating a natural language-to-RDF query engine to our
a football game. In Gridiron Football, an offensive scheme fball project’s SPARQL-endpoint/RDF triple store. The inte-
can be thought of as an offensive strategic system that a team gration of these two technologies allowed our research team
STIDS 2015 Proceedings Page 86
to interrogate the fball event knowledgebase for a collection
of propositions (i.e., statements that are either true or false)
that may be supported by the facts represented in the triple-
store knowledgebase. Principally, a collection or sequence of
statements would be derived to simulate a strategy-recognition
pattern that matches against the RDF-encoded triples in the
store and provide evidence of certain offensive scheme com-
ponents. Identifying certain offensive components such as Pass
Component (Huddle): No Huddle and Passing Component
(Tempo): Hurry-Up Tempo was fairly simple and straight for-
ward to accomplish. This ease to identify a particular offensive
component was due primarily to completeness of data and
that offensive component being directly available within the
fball knowledgebase to directly support that question. In other
cases, even with the detection of an offensive component or Fig. 6. An illustration of the average percentage of short-to-intermediate-
distance attempts versus intermediate-to-long-distance pass attempts by NFL
combination of offensive components it was not sufficient for teams known to not incorporate a West Coast Offense Strategy.
identifying a more complex offensive strategy. To illustrate that
point, here we focus on a particular type of offensive strategy
called a West Coast Offense strategy. As background domain Our researchers asked the following question: "In the time
knowledge, we offer football passing theory that describes duration of a football game or season is this team using a
the West Coast Offense as the concept of using short passes west Coast Offense strategy as part of its offensive scheme?
to replace some of the running attacks [18]. Moreover, the It follows that the research team proposed the following
short pass receiver is expected to run for good yardage after hypothesis, which is two-fold:
the completion. Therefore, a West Coast Offense strategy is a • In regard to the Passing Component (Setup Mode), if the
composite strategy of at least two of the components identified team demonstrates a higher percentage of Pass to Setup
in the previous enumeration: the Passing Component (Setup the Run attempts (i.e., passing attempts) than a team that
Mode) and Passing Component (Length of Passes). demonstrates a higher percentage of Run to Setup the
Pass attempts (i.e., rushing attempts); and
• in regard to the Passing Component (Length of Passes),
if a team demonstrates a higher percentage of Short to
Intermediate pass attempts as compared to the percentage
of Vertical Intermediate to Deep Passing Game passes.
In the evaluation of our detection pattern for the use of a
West Coast Offense strategy in the base offensive scheme for a
particular team, we were not able to identify teams that were
definitely employing this strategy. In figure 5, we illustrate
the average percentage of short-to-intermediate pass attempts
and intermediate-to-long pass attempts that were executed by
NFL teams that were known to use a West Coast Offense
strategy as part of their base offensive scheme during the 2013
game season. In figure 6, the same statistics are illustrated;
however, in this case we show the statistics of teams that were
Fig. 5. An illustration of the average percentage of short-to-intermediate- not known to use a West Coast Offense strategy during the
distance pass attempts versus intermediate-to-long-distance pass attempts by 2013 Season. A quick visual examination of these bar charts
NFL teams known to incorporate a West Coast Offense Strategy.
show that there is not any major difference in the percentage of
short- to long-distance pass attempts between the two category
In figure 7, we show an example of the type of natural of teams (i.e., those known to use a West Coast Offense and
language query used by our researchers for detecting pass those that do not). In addition to the statistics on Passing
completions by a particular team during a single football Component (Length of Passes), statistics related to the Passing
season where the receiver or pass target caught the pass within Component (Setup Mode) also did not show a major difference
5 yards of the goal line and net gained 15 yards or greater between teams known to use a West Coast Offense versus those
after the pass completion. It follows that this particular team, that are not known to use that type of strategy.
which is the 2013 Miami Dolphins, coached by Joe Philbin
and offensively coordinated by Mike Sherman, was known for V. P RELIMINARY R ESULTS & D ISCUSSION
using a West Coast Offensive strategy during the 2013 game A number of insights were made as a result of our
season [19]. In figure 8, we show the results of the query. research effort. First, we believe that the complexity and
STIDS 2015 Proceedings Page 87
Fig. 7. An illustration of a natural language query using the SPARKLIS engine/interface that models the pattern of West Coast Offense scheme .
Fig. 8. An illustration of the query results from the West Coast Offense pattern-based query. Note the illustration shows observations of the West Coast
Offense strategy being used by the Miami Dolphins football team during a game event in 2013
fine-granularity of the real-world, NFL Play-by-Play data set that the modeling of other applications or domains involving
provided our team with an intermediate step for evaluating the human or autonomous agents competing or cooperating within
application of semantic-based event models to the observations a cyber-physical system would present a similar challenge.
of human or autonomous agents engaged in competition at The primary strength of the ontology-based activity (strat-
the scale we would expect within a cyber-physical system. egy) recognition approach is the relative simplicity and
As stated previously, NFL football players and stadiums are straightforwardness involved in incorporating domain knowl-
equipped with RFID sensors and we assert that the type of edge and heuristics into the recognition models. The use
sensor data that we expect to be derived by an NFL Stadium of event semantics was especially beneficial in this regard.
Venue’s cyber-physical system (CPS) will be similar to the For instance, the The Event Ontology provides the ability to
data used by our research team. This type of activity data model and interrogate the NFL event knowledgebase based
provided the research team with a unique challenge in effort on five dimensions: event type, time, location, factor and
to properly reflect in each RDF statement the appropriate outcome. Additionally, the upper ontology also provided an
semantic using the Gridiron Football Ontology. We expect abstraction (i.e., event:hasSubEvent) for describing an event
STIDS 2015 Proceedings Page 88
that is composed of other events. This abstraction allowed the also like to acknowledge Arm Chair Analysis as the source of
research team identify what type of football sub-events were the NFL data set used in this research project.
associated with a particular play event, drive event, game event
R EFERENCES
or season event. For example, a particular play event may be
comprised of a penalty event and a pass event. [1] K. Namuduri, Y. Wan, M. Gomathisankaran, and R. Pendse, “Airborne
network: a cyber-physical system perspective,” in Proceedings of the first
Though challenging, the research team determined that ACM MobiHoc workshop on Airborne Networks and Communications.
developing semantic queries that can detect certain strategies ACM, 2012, pp. 55–60.
being used by NFL competing agents is possible. Initially, [2] T. Olavsrud, “The internet of things comes to the nfl,” Jul 2015.
[3] S. Trimble, “Google targets low-cost ads-b out avionics market,” 2015.
the research team exclusively relied on the development of [4] J. Durham, Football’s Matchup Zone Cover-
SPARQL queries. Thereafter, the team learned that the guid- ages. Harding Press, 1998. [Online]. Available:
ance of an expressive Natural Language-to-RDF query builder https://books.google.com/books?id=qhcKAAAACAAJ
[5] D. Jennings, M. Bahr, R. Danmeier, and D. Herbst, The Art of Place-
such as SPARKLIS is useful for formulating straightforward kicking and Punting. Linden Press/S&S, 1985. [Online]. Available:
queries for answering particularly complex hypotheses. The https://books.google.com/books?id=Z31YAAAAYAAJ
weakness of solely using the ontology-based strategy recog- [6] J. Rice, Defending the Spread Offense. Coaches Choice, 2010. [Online].
Available: https://books.google.com/books?id=lOkXRAAACAAJ
nition approach is the lack of learning ability in terms of [7] A. F. C. Association, Offensive Football Strate-
identifying patterns that can identify certain complex strate- gies. Human Kinetics, 2000. [Online]. Available:
gies. In this case, machine learning techniques for performing https://books.google.com/books?id=pHbvMbMZKjQC
[8] B. Ramseyer, Winning Football. Human Kinetics 1. [Online]. Available:
statistical and probabilistic reasoning may have been useful. https://books.google.com/books?id=IuxRAgAAQBAJ
However, the logical model approach (i.e., ontology-based [9] A. J. Cañas, G. Hill, R. Carff, N. Suri, J. Lott, T. Eskridge, G. Gómez,
activity recognition) can certainly play a dominant role when M. Arroyo, and R. Carvajal, “Cmaptools: A knowledge modeling and
sharing environment,” in Concept maps: Theory, methodology, tech-
it is integrated along with techniques for learning patterns nology. Proceedings of the first international conference on concept
as well as dealing with the inability of the logical model to mapping, vol. 1, 2004, pp. 125–133.
represent fuzziness and uncertainty. We offer that our approach [10] N. F. Noy, M. Crubézy, R. W. Fergerson, H. Knublauch, S. W. Tu,
J. Vendetti, M. A. Musen et al., “Protege-2000: an open-source ontology-
and contribution is an intermediate step that can be further development and knowledge-acquisition environment,” in AMIA Annu
extended to include using an instance of an event ontology Symp Proc, vol. 953, 2003, p. 953.
as a seed ontology for statistical and probabilistic strategy [11] Y. Raimond and S. Abdallah, “The event ontology,” Technical report,
2007. http://motools. sourceforge. net/event, Tech. Rep., 2007.
recognition. In some cases, this seed ontology may be used [12] J. R. Hobbs and F. Pan, “Time ontology in owl, w3c working draft 27
to develop a more comprehensive ontology using ontology september 2006,” W3C Working Draft, vol. 27, 2006.
learning techniques. [13] D. Brickley, “Basic geo (wgs84 lat/long) vocabulary, 2006,” Cité en,
p. 52.
[14] A. Jena, “A free and open source java framework for building semantic
VI. C ONCLUSION AND F UTURE W ORK web and linked data applications,” URL: http://jena. apache. org, 2011.
[15] A. Jena-Fuseki, “serving rdf data over http,” 2014.
In this paper, an approach is given for capability that uses [16] S. Ferré, “Sparklis: a sparql endpoint explorer for expressive question
event semantics to represent the temporal, spatial, factor, and answering,” in ISWC Posters & Demonstrations Track.
[17] G. Morris, “Football 101: Understanding ba-
outcome characteristics of events generated from the observa- sic nfl offensive concepts,” 2014. [Online]. Avail-
tions of agents engaged in a competitive activity between each able: http://www.bloggingtheboys.com/2014/8/11/5965033/football-
other. Further, we have described the likeness of the data set 101-understanding-basic-nfl-offensive-concepts
[18] T. Flores and B. O’Connor, Coaching Foot-
used for this experiment with the kind of data set we would ball. McGraw-Hill Education, 2006. [Online]. Available:
expect to be generated from the type of "cyber-physical game" https://books.google.com/books?id=mrsIBAAAQBAJ
scenario identified in [2]. The approach extends existing mod- [19] W. C. offense, “West coast offense — wikipedia, the free
encyclopedia,” 2015, online accessed 02-September-2015. [Online].
ular vocabularies and is based in the specification of event se- Available: https://en.wikipedia.org/wiki/WestCoastoffense
mantics for our contributed Gridiron Football Ontology using
Resource Description Framework Schema (RDFS) language
and Ontology Web Language (OWL). Our future work has
already begun and includes: extending the event ontology to an
applicable data set for autonomous physical agents cooperating
or competing within a cyber-physical system. Moreover, our
effort seeks to integrate aspects of game theory analysis with
the ontology-based strategy recognition approach to account
for concepts such as payoffs and tensions between the different
strategies that may be used by agents.
ACKNOWLEDGMENT
The authors would like to acknowledge the Sparklis research
activity under Dr. Sebastien Ferre supported at IRISA, Uni-
versite de Remes, Rennes cedex, France. The authors would
STIDS 2015 Proceedings Page 89