Conference PaperPDF Available

Computational Intelligence and Tower Defence Games


Abstract and Figures

The aim of this paper is to introduce the use of Tower Defence (TD) games in Computational Intelligence (CI) research. We show how TD games can provide an important test-bed for the often under-represented casual games research area. Additionally, the use of CI in the TD games has the potential to create a more interesting, interactive and ongoing game experience for casual gamers. We present a definition of the current state and development of TD games, and include a classification of TD game components. We then describe some potential ways CI can be used to augment the TD experience. Finally, a prototype TD game based on experience driven procedural content generation is presented.
Content may be subject to copyright.
Computational Intelligence and Tower Defence Games
Phillipa Avery
Department of Computer Science and Engineering
University of Nevada, Reno, USA
Julian Togelius, Elvis Alistar,
and Robert Pieter van Leeuwen
Center for Computer Games Research
IT University of Copenhagen, Denmark
Email: {juto, elal, rpvl}
Abstract The aim of this paper is to introduce the use of
Tower Defence (TD) games in Computational Intelligence (CI)
research. We show how TD games can provide an important
test-bed for the often under-represented casual games research
area. Additionally, the use of CI in the TD games has the
potential to create a more interesting, interactive and ongoing
game experience for casual gamers. We present a definition of
the current state and development of TD games, and include
a classification of TD game components. We then describe
some potential ways CI can be used to augment the TD
experience. Finally, a prototype TD game based on experience-
driven procedural content generation is presented.
Tower Defence (TD) games are a genre of strategy games
focusing on resource allocation and unit (tower) placement.
In it’s simplest form, a TD game consists of a human player
buying and organizing defensive towers that fire upon a set
deployment of different types of offensive enemies (creeps).
For each creep destroyed by the towers, the player earns
resources corresponding to the difficulty of the creep. If
enough of the creeps are destroyed, the player wins the
round. If however enough creeps reach the end of the player’s
map, the creeps win. The tower deployment usually involves
buying, placing and upgrading towers that fire automatically
on the forces.
TD games have proved to be a challenging, addictive and
fun way to pass the time. The simplicity of gameplay and
the challenging tactics combined with the wide variety of
games available in this genre has created a large population
of casual TD gamers. It is this simplicity and availability
that makes TD games a great test bed for Computational
Intelligence (CI) research. Many of the popular video games
used in CI game research are complex in nature, and are
broken down into different components for the purpose of
research. This often creates a less enjoyable game experience,
and is not always a true depiction of the inherent game. Due
to their popularity, TD games provide a large user-base with a
simple game mechanism, and a range of interesting potential
research possibilities. TD games are relatively easy to pro-
gram being computationally and graphically simple. The TD
genre is simple enough to implement completely for research
purposes, while still providing enough of a challenge for
user interest and complexity for test-bed purposes. Many TD
games have tactical and strategic depth, meaning that there
are real challenges to be solved and the results can have
bearing on AI research in general. Further, there are clear
possibilities to improve the gameplay of TD games using CI
TD games can be real-time or turn-based in nature, and the
allocation of towers allows resource allocation and strategic
deployment. Additionally the choice and timing of ’creep’
deployment also includes resource allocation and scheduling
areas of research. The game can also be played in single
player, multi-player (competitive and cooperative), and as
described later in this paper, has the potential for a zero-sum
two player game. All of these topics are popular problems
to tackle using CI techniques, and the TD genre can provide
a coherent test-bed. The popular use of online play for TD
games also provides a potentially very large set of users to
test with.
This paper defines the TD game for the purpose of future
research, including all the sub-genres and off-shoots from
popular TD games. The taxonomy of sub-genres provides
increased complexity and interest, and can itself be a point
of research potential with procedural content generation. We
provide a brief history of the TD genre, including some
of the most popular TD games. We discuss the possible
reasons for why different games have become popular, and
how future games created through research could cross into
public popularity.
After providing the structure of TD games, we identify
areas where improvements could be made to the game using
CI techniques. We discuss a number of potential fields
of interest, and how the use of TD as a test-bed could
improve the field of expertise. Following this discussion,
we describe a simple TD prototype that has been developed
to demonstrate using CI techniques in the TD genre, and
the results of some first experiments in evolving towers and
creeps for this game.
A. An Example
As a simple example of a TD game, we provide a descrip-
tion of the Flash Element TD [1] that was inspired by the
Element TD for Warcraft III (one of the first implementations
of TD games). Figure 1 provides a screen shot of the game.
This game consists of three different tower types: one that
has low fire power but targets both land and air creeps, and
two that target either land or air creeps respectively. As you
earn more Gold, you can buy more expensive but stronger
towers, and upgrade your existing towers. In the figure, three
towers have been placed on the map grid. In this version of
Fig. 1. Example screen shot of the Flash Element TD game.
the game, towers can only be placed on the darker areas
shown in green. The creeps follow the lighter brown path
from the start to the end position. For each creep that reaches
the end, a life is lost and the creep re-enters the map at the
start position. The round (level) finishes when all creeps have
been destroyed, and this game has a total of 31 levels with
increasingly difficult creep deployment.
Due to the gentle learning curve and ease of playing in
small chunks, this game genre is very popular with “casual”
gamers; such players tend to play games that require little
time investment and offer almost instant progress, and often
do not think of themselves as “gamers” [2]. However, there is
scope within the genre for “hardcore” play as well, as levels
and game modes can be constructed that require significant
strategic sophistication to win.
B. History of TD Games
The TD genre is a fairly recent one, gaining popularity
particularly due to the online flash versions available for
casual gamers. Arguably the first commercial TD game was
the 1990 Atari arcade game Rampant. Due to its popularity,
Rampant was subsequently released on other platforms, and
went on to offer a multi-player version on the Playstation
Network in 2007 [3], [4]. Although Rampart is not the typical
TD game, it does involve defending a set of coastal castles
by shooting attacking sea forces.
Following Rampant, a number of popular real-time and
turn-based strategy games included TD style additions to
the game. One of the most popular versions of these, and
arguably the original true TD style game was the Tower
Defense maps for the Warcraft III expansion The Frozen
Throne. Using the map-making abilities of Starcraft and
Warcraft III, players would create defensive maps similar
to what we now know as TD games. When the expansion
came out, a secret level held a TD game that demonstrated
many of the TD elements seen in the classic version of the
game [5]. The Warcraft III TD also allowed for multi-player
game-play, where players compete to kill the most amount
of creeps. This early link between Real-Time Strategy (RTS)
and TD games highlights the similarities between the genres,
which can be used to a researcher’s advantage.
TD games then grew in popularity with the use of Flash
web games. There are a large number of popular flash TD
games, and competition to gain players’ attention has created
many variations on the original game. The following section
describes some of these game elements, with reference to
games that made them popular.
C. Elements of Tower Defence Games
The genre of TD games is varied, and there does not really
exist a set of boundaries that define the genre. It is not our
purpose to define a set of boundaries here, but instead to
establish a set of features that TD games have displayed over
the years. These features can be combined to make up the
elements of a TD game. We ostentatiously define the genre
by categorizing the game elements that the majority of TD
games utilize.
1) Terrain: The TD map forms the constraints on how the
player can allocate towers. In the classic TD game, the map
is dominated by a linear path for creeps with surrounding
area for building towers (as shown in Figure 1). This path
can remain unchanged, or more commonly different levels
can display more challenging paths. The map can further
restrict the allocation by only allowing towers to be built
in specific locations, such as specific terrain. One example
of this is GemCraft, where players buy gems that can be
placed in towers. A map begins with some towers created,
and players can then build additional towers in permitting
locations. Restricting usable terrain allows the creation of
more difficult strategic deployment on an otherwise relatively
simple game map.
One of the most popular Flash TD games is the Bloons
TD games by Kaiparasoft. Bloons TD uses simple map de-
sign, with different maps categorized into easy, intermediate,
advanced and expert. The map design has areas of high
grouping impact, where towers can reach creeps (balloons
in this case) for a longer time, and also areas where there
is little impact. The easier maps are linear ones, with lots
of high impact areas. The harder maps include branching
paths, where creeps can randomly choose different paths to
take. Figure 2 shows one of the maps from the Bloons TD
4 game, and is listed as an expert level map.
The Bloons example is typical of one genre of TD games.
The game itself is simple, maps are easy to understand, and
the graphics are pleasing yet uncomplicated. The complexity
in the game comes from the unlocking of new and more in-
teresting towers, and completing the levels with increasingly
harder deployment of creeps.
Other popular games following this form of linear and
branching paths include Vector TD,GemCraft, and many
others. While many games include the path-type map design,
with linear and branching paths, other games are more free-
form. The classic example of this genre is the Desktop TD,
where there is no designated map, and instead the tower
placement forms the path design. Creeps at any instance
attempt to take the shortest path towards the end of the
Fig. 2. Example screen shot of the Bloons TD game.
map, as constrained by the towers. Other games include maps
where the placement of towers are limited (such as to only
land in a land/sea terrain), and creeps can come from any
direction. With this style the goal is not to defend the end
position on a map, but to survive the round with your towers
As a final example, a more recent desktop TD style game is
Plants vs. Zombies game. This game consists of parallel
lanes where different towers (flowers) can be built to defend
against the creeps (zombies). The creeps also follow the same
lanes, and the areas for building towers is limited. In this
version of TD, the creeps can destroy the towers to get to
the end position. Once a single creep reaches the end, the
game is over.
We can therefore distinguish between the following types
of TD maps: linear and branching paths,free-form path,
survival and parallel lanes. Naturally, hybrids between these
are possible, as are entirely different forms of terrain.
2) Towers: The building of the towers requires strategic
deployment and the allocation of resources. Most games
allow towers that have different capabilities and costs. In
the classic game, different towers target specific types of
creeps (e.g. the land and air creeps as in the Element
game discussed). Towers that can target multiple types of
creeps are generally weaker. Towers also have different firing
ranges and speed,damage capability,attack formations,
and effects. These attributes can be improved through the
purchase of more advanced towers, and through upgrading
existing towers.
Many games specialize in providing a variety of differ-
ent types of Towers. For example, the Bloons TD series
discussed earlier has many towers that perform different
functionality. The simple towers fire a single linear attack
at a limited time interval. Other towers fire multiple attacks,
attacks in different directions, attacks that damage multiple
creeps at once, and ‘affect’ attacks that both damage and
slow or stop the creeps. Later versions of the game also
included towers that can locate hidden creeps, towers that
gain increased resources and ones that can send creeps back
to the starting position.
Most recent games either provide a large variety of towers
such as Bloons and Plants vs. Zombies, or allow different
functionality through changes to the towers. For example, the
GemCraft TD game allows players to assign different gems
to towers. The color of the gem determines the functionality
of the tower, and the gems can be moved and combined to
perform differently. The functionality includes such effects
as slowing creeps, reducing the armour of the creeps, dam-
aging multiple creeps at once (either in a splash or chain
formation), damage over time attacks, damage multiplier,
and generation of higher resources. Gems can then be com-
bined for upgrade capabilities. Combining gems of the same
colour provides towers with high specialized powers, while
combining gems of different colours diminishes the power
of the functionality, but allows a single tower to perform
multiple functionalities. The gems can then be swapped and
moved (with a loading penalty) among the existing towers,
and new towers created to deploy the gems. This increases
the adaptability of strategies during the real-time game play,
and the combination of gems also adds increased interest.
Another game that provides combination of effects well is
the Onslaught TD game. In this game, there are four main
types of towers. These towers can be individually built and
upgraded, but when they are placed in a specific formation
they form combination attacks. For example, if a red tower
fires on a creep first, followed by a blue tower, then this will
generate a “Lazer Rocket” combo which is stronger than
both towers. As the combo is created by whichever tower
hits first, the range and where the tower is built both come
into play. Onslaught 2 has a total of 20 combo attacks, each
of which has a different graphic and affect.
Other games provide a ‘hero’ where you can choose a
particular type of movable unit that has varying capabilities.
This unit can normally be directed to a particular task,
including directing them to attack a particular creep.
3) Creeps: The types of creeps used by the game are
also directly influenced by the types of towers, and require a
careful balance. Generally creeps consist of types matching
the towers, for example air and ground creeps, or matching
damage types such as ice, fire etc. They also have different
abilities usually consisting of speed,hitpoints and armour
strength. Slow creeps are normally harder to kill, while fast
creeps have fewer hitpoints. The armour ability can be linked
to a specific tower ability - such as a certain tower being able
to lower or destroy a creep armour, thus making it easier
to kill. The more difficult the creep is to kill (using any
combination of the mentioned abilities), the more resources
are generally obtained for killing it.
Most TD games also have a form of ‘boss’ creep, which
are particularly hard to kill. These bosses tend to be deployed
at the end of a round (when resources might be depleted),
or on their own level. Generally if a boss reaches the end of
the map, then the game is over.
4) Reward Systems: Most of the TD games are single
player with a human playing against a static deployment of
creeps. Once the particular deployment of creeps have been
mastered, new challenges are needed (or the player moves
onto another game). To increase the interest and longevity
of TD games, different types of experience structures and
achievement systems have been introduced. For example, the
GemCraft series of games allows players to gain experience
when a map is defeated, and can then use this experience to
upgrade their overall abilities, such as increase all damage
by a particular tower type. The TD game Cursed Treasure:
Don’t Touch My Gems! has a skill set that matches the
three different types of towers available. As more skill points
are earned and spent, different abilities are increased and
unlocked. The experience structure encourages players to
play levels more than once, to increase their abilities to a
point where the next level is achievable. It also adds a new
challenge mechanism to the game-play.
Other games and TD hosting websites have introduced
achievement structures. These allow players to ‘tick off
certain achievements gained through the game. These might
include such things as number of creeps killed, number of
gems combined, towers built etc. As well as challenging
users, the achievements also encourage players to repeatedly
play the game to achieve the harder achievement rankings.
5) Single or Multi-player: Most of the TD games are
single player games, with a human player competing against
a static set of creeps. Sometimes the creeps have some
form of random deployment in timing or types, however this
is restricted to maintain balance of the game. Multi-player
versions involve players working in a cooperative way to beat
the creeps, but competing to obtain the highest kill-rate.
D. Strategies
The strategies involved in TD games can be varied de-
pending on the elements of the game. Different types of
creeps require different strategies, and a well-designed game
will be balanced to be challenging. Overall however, the
strategies can be grouped into two main problem domains:
resource allocation and unit placement. The resource al-
location strategies usually involve choices between buying
lots of cheaper towers and upgrades, or saving up for more
expensive towers/upgrades. While it is common to buy cheap
towers early on, one useful strategy is to buy minimal cheap
towers and save for a powerful tower to place at a strategic
location on the map. Other games that have different types
of creeps often have creeps that come less regularly (such as
air units). It can be tempting to forget to allocate resources to
towers that target these units, and not have enough to counter
waves of these creeps when they appear.
The placement of units is fully dependent on the map and
the creeps being played against. While it is useful to have a
strong front-line, and destroy creeps as they are deployed at
the start position, fast creeps can potentially run straight past
the front-line and additional towers further along the path are
needed to stop this.
From the perspective of a game developer, it is necessary
to balance all the elements chosen for the game with the
creep deployment. For example, if there are towers with
different abilities such as an area affect or slowing etc, the
creeps should be deployed in such a manner to encourage
use of these abilities. Having a range of creeps with different
strength and speed is essential, and defining how to deploy
these creeps to offer the most challenging experience for the
user is an interesting balance problem.
We have described the basic TD game and the various ele-
ments that can make up a TD game, and now we discuss how
the various research areas within computational intelligence
are applicable. We begin with a discussion on the general
usefulness of the TD genre to the academic community, and
then offer specific areas that could benefit from a TD test-bed
A. Benefits of a TD research game
As mentioned previously, the TD game genre provides an
opportunity for creating simple games that can be utilized for
research purposes while being placed in the public domain.
There are at least two good reasons to combine computational
intelligence techniques with TD games: to find useful, fair
and accessible benchmarks for comparing the performance
of CI algorithms, and to find ways to improve the design
and development of TD games, for example by enabling
new game types or streamlining the development process. (In
other words, the same motivations exist here as for research
in CI and games in general.)
Regardless of the motivation, CI techniques could be
applied to TD games in a number of different roles. Below,
we attempt to survey the various roles CI could play in a
TD game, pointing out related research that has been done
on games from other genres.
B. Map Generation
The map is one of the core features of a TD game, and one
of the main drivers of challenge and sources of differentiation
between levels. One of the most common reasons a player
stops playing a TD game is that there are no more levels to
discover. In light of this, automatic map generation would be
a major advantage for a TD game. Map generation could be
done both offline, when the game is developed, or online, on
the individual player’s computer in response to the playing
style and preferences of that player.
Map generation could be done in more or less simplistic
fashions, for example through starting with a straight path
and changing direction arbitrarily, or randomly placing ob-
jects on an empty surface and find the shortest path using
a pathfinding algorithm. However, such methods are not
very controllable, meaning that there is no good way of
ensuring that the generated map is of the right difficulty,
that it promotes the right kinds of towers and strategies, or
even that it is playable.
A CI-based alternative is the search-based paradigm,
where a stochastic optimization algorithm such as an evo-
lutionary algorithm is used to search a content space for
game content that maximizes certain criteria [6]. In case
of a tower defence game, one would start with finding a
representation for the map, and an operationalization of some
desirable criteria for that map. These criteria could be that
the map should have the right level of challenge, that it
should promote combinations between certain towers, force
a particular strategy etc. Once these criteria are formalized
as a fitness function, maps can be evolved that satisfy them
as well as possible given the game and map representation.
The search-based approach has previously been used to
evolve maps for the real-time strategy game StarCraft [7].
In this experiment, the bases and resources on the maps
were represented directly as (x, y)positions, whereas the
rock formations that serve to separate parts of the map from
each other were represented indirectly as parameters for a
turtle-like (strongly defensive) mechanism. Several fitness
functions, having to do with the fairness of the maps and their
suitability for using advanced strategies (e.g. the presence of
choke points) were defined, and combined using a multi-
objective evolutionary algorithm. A similar approach could
conceivably work well for TD games, once the appropriate
map representation and evaluation functions are defined.
C. AI for playing the game
The deployment of resources in Real-Time Strategy (RTS)
games has been a hot research topic in the last few years,
with research performed on different ways to deploy units
to on a map. Miles, Avery and Louis [8], [9] co-evolve the
strategic deployment of units on an RTS grid using Influence
Maps. Other research such as that by Weber and Mateas [10]
focus on methods to automate the build order (when to build
specific types of units); Hagelb¨
ack and Johansson use multi-
agent techniques and potential fields for controlling the units
in RTS games [11]. These problems are also relevant to TD
games, where the pertinent questions are where to build what
towers and when. The TD map is also usually represented in
a grid format similar to RTS games, so successful techniques
for TD games might also be applicable for RTS and similar
turn-based games.
The resource allocation area is another interesting research
topic. Research on using CI techniques for a turn-based
resource allocation game has been done by Johnson et. al.
[12] and Avery and Michalewicz [13]. As a large component
of the TD strategy is through resource allocation, it would
be interesting to see if similar techniques could be applied
Also, the strategies involved for playing TD games can
be varied, and success can be dependent on how and when
creeps are deployed. While there is often a ’winning’ solu-
tion, where a given strategy will always beat a given set of
creeps, a more dynamic deployment of creeps could create
some interesting strategy development. This concept lends
itself particularly well to coevolution, where both successful
creep deployment and tower strategies could be discovered
for human game play. This is discussed further in the next
D. Creep strategy
One possible change to the TD game that the CI commu-
nity can make, is to make the deployment of creeps dynamic
or even dependent on the player’s strategy. When using a
dynamic creep deployment, agents could be created to play
both sides of the TD game. This would require a method
of restricting the way creeps can be deployed. One possible
method of this is to treat creeps in a similar way to towers,
in that the creep “player” earns resources by outlasting the
towers. The further a creep progresses along the map path,
the more resources earnt. These resources can then be used to
purchase more creeps, with higher costs for stronger, faster
and tougher creeps. It is then up to the creep player to
determine what creeps to buy and when to deploy them in
order to survive the opposition player’s tower placement.
By changing the game in this manner, it is also possible
for human players to play both sides of the game, as either
creeps or towers. This provides many interesting research
possibilities, as well as a more challenging game for human
players. Play could either be done in real time, or turn based.
E. Game element generation
In addition to automatic map generation, there also exists
potential to change the dynamics of the game itself. As
discussed in Section I-C, there are a number of different
variations on the TD theme. Some of these are easily
exchangeable, such as changing the attributes towers (and
corresponding creeps) are allowed to have. Finding a way to
automatically choose and then balance these attributes could
create another method of continuing game-play for a TD
game. Demonstrating the ability for CI techniques to find
continual balanced and engaging gameplay would also show
the usefulness of CI techniques at doing this for commercial
game prospects. The generation of elements and game AI
could also be combined to automatically adjust the difficulty
of the game. This is discussed further in the next section.
F. Dynamic difficulty adjustment and Player Modelling
Game balancing, or dynamic difficulty adjustment (DDA),
refers to the automatic adjustment of games to fit the skill
level of the player [14]. This is motivated by the ever-
widening demographic of game players, and ever-increasing
costs of developing top-quality games. Simply put, a single
game will need to appeal to a much larger range of different
players and different playing styles now than it did back
in the days when most game-players were teenage western
males with a technology interest (and most games conse-
quently were brutally punishing). The DDA answer to this
challenge is to estimate the playing skill of the player and
adapt the game in real-time so as to provide a reasonable
level of challenge — not too hard, not too easy. Sometimes
referred to as “rubber-banding”, DDA has been criticized
for removing the dramatic profile of games, and rewarding
mediocre gameplay. Still, it is a commonly used technique,
and how best to balance challenge is an ongoing research
topic. It could be argued that the less obvious the adjustments
are to the player, the better.
While difficulty adjustment could be implemented in a
simplistic manner in a TD game, e.g. by lowering the health
or speed of incoming creeps if the player is doing badly,
more effective and less obvious methods could rely on map
generation or creep strategy generation as discussed above.
If the player is doing too well or too badly, next level’s map
could be evolved to be easier to finish, or the next creep
wave could be evolved to be easier to beat given the existing
weaponry. If the next level and wave will be dynamically
generated regardless of the player’s performance, this will
make the adjustments hard to detect to the player. A similar
suggestion for platform games is made in [15].
Another method of adjusting the difficulty is to create a
model of the player’s strategy, and develop a correspond-
ing strategy specifically for that player. Work by Avery
and Michalewicz [16] follows this method, and showed
the potential for a fun and challenging game. Using this
method in TD games could increase the enjoyment, when
players know the game they are playing is tailored to their
game playing experience. The grid nature of the TD game
also gives potential for some interesting player modelling
techniques, such as classification of tower placement and
using th prediction techniques.
A further refinement that CI techniques make possible
is to step away from a one-dimensional view of player
performance, and instead see performance as having sev-
eral components and adaptation to these as positioning the
challenge profile in a multidimensional space. For example,
a particular player might be good at placing the offensive
towers for maximum effect, not very adept at using the sup-
porting towers, make good investment decisions (i.e. when
to buy towers and when to save money and get interest), be
mediocre at selecting which gates to protect and have slow
reactions to changing circumstances. These relative strengths
and weaknesses can be measured and used to inform a
fitness function, so that content can be evolved that plays the
unique strengths and weaknesses of the player. This way, an
appropriate overall different level can be maintained at the
same time as the player is forced to explore new strategies
and tower types.
In order to further personalize the game, the content
generation can be made experience-driven [17]. This means
adapting the game not only to the skills of players, but
also to their preferences. For this, player experience models
need to be created, mapping features of the game and
gameplay to predicted player experience. Such models can
be created through administering preference questionnaires
after gameplay sessions, and mapping recorded gameplay
features to the expressed questionnaires using neuroevolu-
tion [18]. This technique has previously been used to create
neural network models that accurately classify levels in the
classic Super Mario Bros platform game in several different
affective dimensions (fun, frustration, interestingness etc.) for
Fig. 3. The Infinite TD game at the start of a level, before any towers have
been placed and any creeps appeared. The 4×4grid at the top of the screen
is the tower repository, from which towers can be bought for placement on
the map (or discarded).
individual players, allowing the creation of levels that elicit
desired affects in particular players [19].
Infinite TD is a prototype tower defence game aiming to
demonstrate how CI and other adaptive techniques can be
used to improve this genre of games. Instead of taking an
existing TD game and modifying it, the decision was made
to design the game from the ground up around adaptive
mechanisms. This section describes the overall design and
adaptive features at the current state of the game; the game
is currently being developed, with a target of being released
on major game platforms (e.g. the iPhone app store) late
summer 2011.
A. Game Design
In some senses, Infinite TD is an utterly conventional
tower defence game. The goal is to survive as many waves of
creeps as possible. Every nwaves, a new path is generated,
and the difficulty of the new path is adjusted to counter-
balance the player’s performance. The creeps travel along
the path from beginning to end without any possibilities for
branching of diversion, and for each creep that reaches the
end of the path, the player loses a life. To defend against
the creeps, the player can purchase towers and place them
strategically on the sides of the path. The price of a tower
depends on its capabilities, which can vary along dimensions
such as range, firing speed, poison effect etc., and money
to purchase more towers is earned by killing creeps. Two
screenshots of the game are shown in figures 3 and 4; the
captions of those figures explain the elements of the game’s
user interface.
The non-conventional aspects of Infinite TD fall into
three categories: map generation, creep evolution and tower
Fig. 4. The Infinite TD game during play of another level, in the middle of
a creep wave. The purple stars inside the dark path are creeps, moving from
the dark green path entrance towards the red path exit. Above each critter
is a small green bar signifying its current health. The brightly coloured
pentagons outside of the path are towers. One of the towers is currently
selected, meaning that a dimly illuminated circle around the tower shows
its range, and the info box to the right of the screen displays its statistics
(range, speed, splash range etc).
B. Map Generation
An Infinite TD level is easier if the creep path is long and
winding, and harder if the path is short and straight. One
reason for this is that a long path gives the players more
time to react to the creep wave, but the main reason is that
a winding path effectively increases the range of towers, as
they can be placed so that they cover larger parts of the path.
Each time the level of difficulty changes, a new map is
generated based on the two parameters linearity and length,
using a simple constructive approach. The algorithm selects
one spawn point on a random edge of a map and a target
point on another random edge of the map. The algorithm
starts carving a path between these two points changing
directions depending on the linearity value. If the algorithm
gets stuck (e.g. path closes on itself), it backtracks to a
previous point and changes the direction of carving. This
process repeats until the carved path connects the spawn
point and the target point and the path has the minimum
length defined by the difficulty level.
C. Creep Evolution
At the beginning of a new wave, new creeps are evolved
to be maximally effective against the previous strategy of the
player, so as to force players to keep redefining and refining
their strategies. Each creep wave is defined by the number
of creeps, speed of the creeps, hit points, armor and spawn
delay. When evolving a wave, the chromosome contains four
genes in the range 01; these are used to calculate the creep
wave in combination with a number of points. The number
of points act as a strength parameter for the creep wave,
and can be used to adapt the difficulty of a wave without
changing its composition. The wave is calculated from the
Fig. 5. Creep evolution. The “evolution points” value refers to the number
of points that need to be spent before attaining a fitness of at least 1, meaning
that at least one creep breaks through the player’s defences and reaches the
end of the path. In this game, the creeps are tested against a complicated
map with 30 towers.
genes as follows:
Amount of creeps in a wave: (gene 0) * 20 + 1
Points per creep: (Total number of points) / (Amount of
creeps + 2)
Total weighing factor: (armor gene) + (speed gene) +
(hitpoints gene)
Speed: (speed gene)/(total weighing factor) * (Total
number of points) * (Speed modifier)
Armor: (armor gene)/(total weighing factor) * (points
per creep) * (Armor modifier)
Hitpoints: (hitpoints gene)/(total weighing factor) *
(points per creep) * (hitpoints modifier)
The fitness calculation of a creep wave is simulation-
based: the wave is played out against the player’s previous
tower configuration (on the previous level) and the fitness
value is calculated as the number of creeps that made it
through to the end of the path, or the distance the creeps
managed to travel along the path if no creeps made it to
the end. This fitness is averaged over several playouts, as
the outcome of a single game is slightly nondeterministic. It
should be noted that thousands of waves can be simulated
per second on a normal desktop computer, meaning that it is
entirely feasible to evolve a new wave in just a few seconds.
Figure 5 shows the results of a series of experiments,
where it was investigated how fast good creep wave config-
urations were evolved using a simple genetic algorithm and
different population sizes. A good wave configuration is in
this context one where a smaller number of points is needed
to break through the defences and reach fitness 1 (one creep
gets to the end of the path). The number of points needed
is calculated through playing the wave starting with 2000
points, and if it doesn’t reach at least fitness 1 increasing the
fitness with 5% cumulatively until that fitness is reached.
D. Interactive Tower Evolution
In order to keep the game challenging at all times we
decided to give the players different towers that evolve
according to their preferences. The towers are evolved using
a genetic algorithm, and each individual uses chromosomes
with 11 genes, 3 for the basic Range, Speed and Damage
properties common and necessary for all towers, and 8 for
the optional properties that add bonuses to the towers. The
full list of the genes and their respective ranges is:
Gene 0: Tower range: 50 - 200
Gene 1: Shooting Speed: 1 - 4
Gene 2: Splash Range: 0 - 80
Gene 3: Damage: 3 - 100
Gene 4: Slow: 0 - 80
Gene 5: Resource Increase: 0 - 10
Gene 6: Triple Damage: 0 - 80
Gene 7: Armor Reduce: 0 - 80
Gene 8: Poison: 0 - 50
Gene 9: Shock: 10 - 80
Gene 10: Chain Hit: 0 - 8
Each gene contains a float value between 0 and 1, which
is translated into the specified range. For example, in order
to find out the range of a tower that’s the result of a
value of 0.63 in Gene 0, we used the following formula:
geneV alue (maxRangeV alue minRangeV alue) +
minRangeV alue, which in this case means 0.63 ×(200
50) + 50, resulting in a value of 144.5.
When the game starts the algorithm creates a population
of 16 individuals. Two of these individuals have their genes
initialized with completely random values, while the other 14
individuals have their genes initialized with values between
0 and 0.2. The initial range limitation is in order that the
players don’t get towers that are too powerful in the begin-
ning of the game. Every time the player builds a tower, the
population size increases with one and every time the player
sells a tower, the population size gets decreased by one. The
population size is always equal to the number of towers the
player has on the screen at any time. When the player sells a
tower or destroys one in the tower selection grid, a new tower
needs to be generated. The new tower is generated as the
first child from the one point crossover between two random
individuals in the population. A Gaussian mutation is then
applied to one of its genes. Every 15th tower that the game
generates will be based on an individual with completely
random values in its genes.
This paper has discussed the use of Tower Defence (TD)
games in the CI field. We have given an overview of the
game genre, and discussed the various elements that make up
a TD game. We have also identified a number of interesting
research areas where TD games would be particularly useful
as a test-bed. Finally we introduce the first step in the creation
of a game that investigates some of these research areas.
As shown, there are a wide and varied range of research
possibilities for using the TD genre in CI research. The addi-
tion of CI techniques to the TD game could also create a very
challenging and interesting game with commercially viable
possibilities. In the future we hope to combine these two
possibilities to create a game that the research community
can experiment and learn from, and that the public can enjoy
on an ongoing basis. The use of the game by the public will
also make further data available for research.
Following on from the development of this game, we
would like to create a competition to find even more in-
teresting and exciting ways to extend the basic TD game
using CI techniques. One way of doing this is to have a
subjective competition, where the entries are placed in the
public domain and the most popular (most played, or a voting
system) entry wins. This could be in addition to an objective
competition, where the most high scoring AI wins.
We hope that this paper will spark interest in the use of
TD games in the CI field, and that future work will prove the
importance of this simple yet strategically complex game.
[1] (2011, Jan.) Flash element td game. [Online]. Available: http:
// td/
[2] J. Juul, A Casual Revolution: Reinventing Video Games and Their
Players. MIT Press, 2009.
[3] (2008, Aug.) The history of tower defense games. [Online]. Available: tower-defense-games/
[4] L. Mitchell. (2008, June) Tower defense: Bringing the
genre back. [Online]. Available:
tower-defense-bringing- the-genre-back/
[5] C. Tamas. (2010, Mar.) Understanding tower defense
games. [Online]. Available:
understanding-tower-defense- games/
[6] J. Togelius, G. N. Yannakakis, K. O. Stanley, and C. Browne, “Search-
based procedural content generation,” in Proceedings of EvoApplica-
tions, vol. 6024. Springer LNCS, 2010.
[7] J. Togelius, M. Preuss, N. Beume, S. Wessing, J. Hagelb¨
ack, and G. N.
Yannakakis, “Multiobjective exploration of the starcraft map space,
in Proceedings of the IEEE Conference on Computational Intelligence
and Games (CIG), 2010.
[8] C. Miles and S. J. Louis, “Co-evolving influence map tree based
strategy game players,” in Proceedings of the 2007 IEEE Symposium
on Computational Intelligence in Games. IEEE Press, 2007.
[9] P. M. Avery, S. J. Louis, and B. Avery, “Evolving coordinated spatial
tactics for autonomous agents using influence maps,” in Proceedings
of the 2009 IEEE Symposium on Computational Intelligence in Games.
IEEE Press, 2009.
[10] B. G. Weber and M. Mateas, “Case-based reasoning for build order
in real-time strategy games,” in AIIDE, 2009.
[11] J. Hagelb¨
ack and S. Johansson, “A multiagent potential field-based
bot for real-time strategy games,” International Journal of Computer
Games Technology, vol. 2009, 2009.
[12] R. Johnson, M. Melich, Z. Michalewicz, and M. Schmidt, “Coevolu-
tionary optimization of fuzzy logic intelligence for strategic decision
support,” in IEEE Transactions on Evolutionary Computation, vol. 9,
no. 6. IEEE, 2005, pp. 682–694.
[13] P. Avery, Z. Michalewicz, and M. Schmidt, “A historical population
in a coevolutionary system,” in IEEE Symposium on Computational
Intelligence and Games. IEEE, 2007, pp. 104 – 111.
[14] R. Hunicke and V. Chapman, “AI for dynamic difficulty adjustment
in games,” in Proceedings of Artificial Intelligence and Interactive
Digital Entertainment, 2004.
[15] M. Jennings-Teats, G. Smith, and N. Wardrip-Fruin, “Polymorph:
A model for dynamic level generation,” in Proceedings of Artificial
Intelligence and Interactive Digital Entertainment, 2010.
[16] P. M. Avery and Z. Michalewicz, “Adapting to human gamers using
coevolution,” in Advances in Machine Learning II, ser. Studies in
Computational Intelligence, vol. 263. Springer Berlin / Heidelberg,
2010, pp. 75–100.
[17] G. N. Yannakakis and J. Togelius, “Experience-driven procedural
content generation,” IEEE Transactions on Affective Computing, vol.
in press, 2011.
[18] G. N. Yannakakis, “How to Model and Augment Player Satisfaction:
A Review,” in Proceedings of the 1st Workshop on Child, Computer
and Interaction. Chania, Crete: ACM Press, October 2008.
[19] C. Pedersen, J. Togelius, and G. N. Yannakakis, “Modeling Player Ex-
perience for Content Creation,” IEEE Transactions on Computational
Intelligence and AI in Games, vol. 2, no. 1, pp. 54–67, 2010.
... The type of the Tower Defense genre described in this paper is based on the genre described in [8]. The Tower Defense genre emerged during the time of Flash-Games. ...
... As the game progresses, the opponents become stronger and the player is encouraged to further populate the map with towers. Although there are many different forms of the genre, the following basic elements can be identified in each Tower Defense game (see [8] pp. 1085-1087). ...
Full-text available
This paper analyzes the revenue models of the most popular games of the Tower Defense genre on Google Play. A special look is taken at the quantitative distribution of the app sale model and the free model in terms of quality and download numbers. Additionally, this paper considers the qualitative implementation of the "free" model in the most popular games. First, the usual revenue models of mobile apps will be discussed and then the Tower Defense genre will be explained. Following that, the quantitative distribution of revenue models and an analysis of the most popular apps' respective revenue models will be addressed. The analysis also identifies and explains two modifications of established revenue models. The most popular revenue model for mobile apps in the Tower Defense genre are in-app purchases. This distinguishes the genre from many other genres and games. A wide range of Tower Defense games utilizes the revenue models app sale and free. It becomes apparent that revenue models for mobile apps must be analyzed and considered specifically for their respective sector, and that no single promising revenue model for apps exists.
... Although the use of PCG applied to games has been around since the 80s [13], the amount of scientific work on the use of PCG in TD is not very extensive. One of the first studies on TD games was presented in [14]. In the paper, the authors identify areas where TD games would be particularly useful to intelligent computing. ...
... Science education teaches science to school children and non-scientists, and the gaming industry has encouraged its growth by developing Serious gaming is the use of games to fill needs other than just entertainment [3,4] . Serious games are also called applied games, and this concept of gaming is used by industries like defense [5] , health care [6] , emergency management [7,8] , education [9] , sports [10,11] , exploration [12] , city planning [13] , engineering [14] , and politics [15] . In addition to this, computer vision and deep learning-based techniques can be exploited in a gaming context to analyze performances of sports players through tracking [16][17][18][19] in a virtual environment [20,21] , detection [22][23][24] , analyzing anomalous behavior [25][26][27] , simulating individual [28,29] and crowd behavior [30][31][32][33][34][35] for public infrastructure design [36,37] , and a variety of other multimedia applications [38][39][40] . ...
Full-text available
Teaching science through computer games, simulations, and artificial intelligence (AI) is an increasingly active research field. To this end, we conducted a systematic literature review on serious games for science education to reveal research trends and patterns. We discussed the role of Virtual Reality (VR), AI, and Augmented Reality (AR) games in teaching science subjects like physics. Specifically, we covered the research spanning between 2011 and 2021, investigated country-wise concentration and most common evaluation methods, and discussed the positive and negative aspects of serious games in science education in particular and attitudes towards the use of serious games in education in general.
... The Tower Defense game characterizes as a strategy computer game and was first introduced in the 1990s [1]. Until now, researchers applied PCG to the Tower Defense Game genre, but they either used PCG to generate levels [13], generate paths and monster sequences [4], or generate road maps, tower locations, and monster sequences [9]. ...
Full-text available
In the present day, it is difficult to imagine the development of computer games without the use of artificial intelligence. We see it utilized for gameplay, players modeling, playtesting, or content generation. In this paper, we focused on the content generation of a custom Tower Defense game named Save the Sheep. The Tower defense game is a strategic game, which was, in our case, proposed in a non-violent way. We generated key building blocks of the game with a genetic algorithm, i.e., a game map, unit placement, and a waves system. The final Tower Defense inspired game was implemented in the Unity game engine. The results show that by applying genetic algorithms, it is possible not only to generate content that makes the game more complex, but also more challenging and interesting for players.
... Serious gaming is the use of games to fill needs other than just entertainment [2]. Serious games are also called applied games and this concept of gaming is used by industries like defense [3], health care [4], emergency management [5], [6], education [7], exploration [8], city planning [9], engineering [10], and politics [11]. In addition to this, computer vision and deep learning based techniques can be exploited in gaming context to analyze performances of sports players through tracking [12], [13] in an virtual environment [14], detection [15], [16], analysing anomalous behaviour [17], [18], simulating individual [19], [20] and crowd behaviour [21]- [25] for public infrastructure design [26], [27], and a variety of other multi media applications [28], [29]. ...
Full-text available
Keeping in mind the increasing trend and need for serious games in science education, we have done a systematic literature review. These papers show the trends and patterns of research carried out in this field from the year 2011 to 2020. Specifically, we investigated country-wise concentration and the most common evaluation methods. Literature is reviewed from IEEEexplore, Springer, and Scopus. Moreover, we discussed the role of Augmented Reality(AR) games in teaching physics. Lastly, we have discussed the positive and negative aspects of serious games in science education in particular, and the trend of using serious games in the past decade in education in general.
... We use a tower defense (TD) game developed in-house as our test subject (see Figure 1), although our general approach can be extended to other single-player video games as well. Tower defense is a popular genre for casual gamers and is a proven testbed for research into AI and games [2]. For example, [1] investigates a genetic algorithm to evolve increasingly difficult monster waves for a TD game, [22] studies how to adjust a TD game's parameters dynamically based on a player's performance in previous levels, and [11] evolves automated controllers for playing a TD game. ...
Conference Paper
Tower defense is a popular subgenre of real-time strategy game requiring detailed level design and difficulty balancing to create an enjoyable player experience. Because level production and testing are both time-consuming and labor-intensive, we propose and implement a framework to automate the process. We first analyze the three main components, or "building blocks", of the popular tower defense game Kingdom Rush: Frontiers (KRF), i.e. road maps, tower locations and monster sequences. We then automatically create new building blocks in the style of the original game, utilizing techniques from Procedural Content Generation (PCG), and assemble them to create new levels. We also add a fourth block: automated testing via a Monte Carlo search algorithm, to ensure the generated content is playable. We focus on KRF because it is a popular video game in the tower defense genre, and highlights some of the challenges of designing appropriate PCG and playtesting algorithms for a commercial video game.
... Galactic Arms Race [6] is a space shooter in which the firing patterns of weapons evolve based on usage frequency. Researchers have also evolved levels for Super Mario Bros [7], various aspects of a tower defense game [8], and maps for a First-Person Shooter [9]. ...
... Based on the definition by [4] and [5], the tower defense genre is characterized by (1) the player build defensive towers to obstruct the enemies (or creeps) from reaching the player's base, (2) the enemies come in waves with increased difficulties, (3) by defeating the enemies, the player will get resources corresponding to the difficulty of the enemy, that can be used to upgrade their towers, and (4) the tower can be bought, placed in the map, upgraded, and fire automatically on the enemy forces. As these only serve as a general characteristic, more gameplay and mechanic variation exist on the market. ...
Full-text available
In this paper, we propose a new style of tower defense game by using the circular placement of the tower. In this game, dubbed Orbital Defense X, the player must protect the earth with a circle of defensive tower circled around it. The game is developed using game development lifecycle from initiation phase through alpha version. To test the quality of the developed game, we measure the game experience using game experience questionnaire. The questionnaire is divided into three: core, social presence, and post-game questionnaire. The results show that male participant gains a significantly higher score than female participants. Overall, the game experience is received better by the male participant in the core and social presence components, with an average score of 2,27 vs 1,92 and 2,15 vs 1,86. However, in post-game experience, the female participant scores higher with 2,15 vs 1,95. The participant that previously played another tower defense game gave score more in the core, social presence, and post-game experience than the participant that haven’t played it before. Their score is 1,93 vs 2,37 in core, 1,67 vs 2,39 in social presence, and 2,07 vs 1,91 in post-game experience. The result also suggests for more attention in the challenge regarding the enemies (variety and difficulty) rather than more complex tower management system. The social ranking also suggested by the participant to improve the overall experience of this game.
... (Unity Technologies, San Francisco, CA). Tower Defense has been deemed a reliable genre for research [21]. A typical TD game features waves of enemies who spawn at a fixed rate. ...
Full-text available
Immersive experiences are typically considered an indicator of successful game design. The ability to maintain the player’s focus and enjoyment in the game lies at the core of game mechanics. In this work, we used a custom virtual reality game aiming to induce flow, boredom and anxiety throughout specific instances in the game. We used self-reports of personality and flow in addition to physiological measures (heart rate variability) as a means of evaluating the game design. Results yielded a consistently high accuracy in the classification of low flow versus high flow conditions across multiple classifiers. Moreover, they suggested that the anticipated model-by-design was not necessarily consistent with the player’s subjective and objective data. Our approach lays promising groundwork for the automatic assessment of game design strategies and may help explain experiential variability across video game players.
... The setting of this study takes place within a VR Tower Defence (TD) game as experienced through GearVR. The reasons for this choice are the low entry requirements of GearVR and the benefits of conducting research in this type of scenario as stated by Avery, Togelius, Alistar & Van Leeuwen (2011). Additionally, the fun factor makes the study's environment suitable for introducing new technologies to users. ...
Full-text available
The field of virtual reality (VR) is getting increasing attention from the scientific community and it is being portrayed by advertisements as the user interface (UI) of the future. This is a fair statement since the prior uses of VR that used to exist only in fiction movies and books, are now widely available in many forms and settings to the public. One of the most interesting outcomes from this technological evolution is that now VR can be experienced through the use of a mobile phone and the addition of some inexpensive means typically in a form of a headset. The combination of the phone's screen as attached to the headset, creates a form of Head Mounted Display (HMD) which can be utilized in order for the user to be immersed within a virtual environment (VE). The argument here is that even if the means to get access to VR are cheap, this should not be the case with the experience as well. On the contrary, the low entry requirements in combination with a high quality experience are the basis for the medium's success and further adoption by the users. More specifically, the capability of utilizing a three dimensional space (3D) should not limit the medium's use on just that but instead, this space should be used in order to offer immersive environments which make the user feel as if he is there. There are many factors that contribute to that result and significant progress has been made to some such as the quality of screen or other hardware parts that allow the user get immersed into the virtual scenery, however, little progress has been made towards the conceptual means that allow the user of better experiencing this VE. Most of the VR applications so far are specifically designed for a single user session. This creates an isolation of the user from any other type of communities which further increases the stigma of VR being a solitary experience. Another issue is the interaction method that is available to users in order to interact with the VE. The use of buttons in most of the available headsets is a counter intuitive method for a person to interact with an environment that wants to be called real. The technological advancements in the field of image processing have resulted in many new methods of interaction and multimodal manipulation within VE and it would be worthy of exploring their effects on the user experience (UX) when used as an interaction method. For these reasons, this thesis used the case of VR games as a setting to study how UX can be enhanced from its current state by introducing a bare hand gesture interaction method and expanding the VR setting in order to host two users in shared VE. Two individual studies were conducted where user feedback was collected in order to describe the effects of this approach in both a qualitative and quantitative manner. As results indicate, by utilizing gesture analysis on a headset equipped with a smartphone, it is possible to offer a natural and engaging solution of VR interaction capable of rich UXs while maintaining a low entry level for the end users. Finally, the addition of another player, significantly affected the experience by influencing the emotional state of the participants in the game and further enforcing their feeling of presence within the VE.
Full-text available
Procedural content generation (PCG) is an increasingly important area of technology within modern human-computer interaction (HCI) design. Personalization of user experience via affective and cognitive modeling, coupled with real-time adjustment of the content according to user needs and preferences are important steps toward effective and meaningful PCG. Games, Web 2.0, interface, and software design are among the most popular applications of automated content generation. The paper provides a taxonomy of PCG algorithms and introduces a framework for PCG driven by computational models of user experience. This approach, which we call Experience-Driven Procedural Content Generation (EDPCG), is generic and applicable to various subareas of HCI. We employ games as an example indicative of rich HCI and complex affect elicitation, and demonstrate the approach's effectiveness via dissimilar successful studies.
Full-text available
Video Games are boring when they are too easy and frustrating when they are too hard. While most single-player games allow players to adjust basic difficulty (easy, medium, hard, insane), their overall level of challenge is often static in the face of individual player input. This lack of flexibility can lead to mismatches between player ability and overall game difficulty. In this paper, we explore the computational and design requirements for a dynamic difficulty adjustment system. We present a probabilistic method (drawn predominantly from Inventory Theory) for representing and reasoning about uncertainty in games. We describe the implementation of these techniques, and discuss how the resulting system can be applied to create flexible interactive experiences that adjust on the fly.
Full-text available
This is a review on approaches for modeling satisfaction per-ceived by users interacting with entertainment systems. Ex-perimental studies with adult and children users of games (screen-based and physical-interactive) are outlined and the most promising approaches for augmenting player satisfac-tion while the game is played (i.e. in real-time) are dis-cussed.
Conference Paper
Full-text available
Recently, a small number of papers have appeared in which the authors implement stochastic search algorithms, such as evolutionary computation, to generate game content, such as levels, rules and weapons. We propose a taxonomy of such approaches, centring on what sort of content is generated, how the content is represented, and how the quality of the content is evaluated. The relation between search-based and other types of procedural content generation is described, as are some of the main research challenges in this new field. The paper ends with some successful examples of this approach.
Full-text available
In this paper, we use computational intelligence techniques to built quantitative models of player experience for a platform game. The models accurately predict certain key affective states of the player based on both gameplay metrics that relate to the actions performed by the player in the game, and on parameters of the level that was played. For the experiments presented here, a version of the classic Super Mario Bros game is enhanced with parameterizable level generation and gameplay metrics collection. Player pairwise preference data is collected using forced choice questionnaires, and the models are trained using this data and neuroevolutionary preference learning of multilayer perceptrons (MLPs). The derived models will be used to optimize design parameters for particular types of player experience, allowing the designer to automatically generate unique levels that induce the desired experience for the player.
We used to think that video games were mostly for young men, but with the success of the Nintendo Wii, and the proliferation of games in browsers, cell phone games, and social games video games changed changed fundamentally in the years from 2000 to 2010. These new casual games are now played by men and women, young and old. Players need not possess an intimate knowledge of video game history or devote weeks or months to play. At the same time, many players of casual games show a dedication and skill that is anything but casual. In A Casual Revolution, Jesper Juul describes this as a reinvention of video games, and of our image of video game players, and explores what this tells us about the players, the games, and their interaction. With this reinvention of video games, the game industry reconnects with a general audience. Many of today's casual game players once enjoyed Pac-Man, Tetris, and other early games, only to drop out when video games became more time-consuming and complex. Juul shows that it is only by understanding what a game requires of players, what players bring to a game, how the game industry works, and how video games have developed historically that we can understand what makes video games fun and why we choose to play (or not to play) them.
No matter how good a computer player is, given enough time human players may learn to adapt to the strategy used, and routinely defeat the computer player. A challenging task is to mimic this human ability to adapt, and create a computer player that can adapt to its opposition’s strategy. By having an adaptive strategy for a computer player, the challenge it provides is ongoing. Additionally, a computer player that adapts specifically to an individual human provides a more personal and tailored game play experience. To address this need we have investigated the creation of such a computer player. By creating a computer player that changes its strategy with influence from the human strategy, we have shown that the holy grail of gaming – an individually tailored gaming experience, is indeed possible. We designed the computer player for the game of TEMPO, a zero sum military planning game. The player was created through a process that reverse engineers the human strategy and uses it to coevolve the computer player.
Conference Paper
We evolve tactical control for entity groups in a naval real-time strategy game. Since tactical maneuvering involves spatial reasoning, our evolutionary algorithm evolves a set of influence maps that help specify an entity's spatial objectives. The entity then uses the A* route finding algorithm to generate waypoints according to the influence map, and follows them to achieve spatial objectives. Using this representation, our evolutionary algorithm quickly evolves increasingly better capture-the-flag tactics on three increasingly difficult maps. These preliminary results indicate (1) the usefulness of our particular influence map encoding for representing spatially resolved tactics and (2) the potential for using co-evolution to generate increasingly complex and competent tactics in our game. More generally, this work represents another step in our ongoing effort to investigate the co-evolution of competent game players in a real-time, continuous, environment that does not assume complete knowledge of the game state.