Content uploaded by Lorenz Cuno Klopfenstein
Author content
All content in this area was uploaded by Lorenz Cuno Klopfenstein on Sep 10, 2018
Content may be subject to copyright.
Balance Trucks: Using Crowd-Sourced Data to
Procedurally-Generate Gameplay within Mobile
Games
Mark Russell Lewis, Sylvester Arnab, Luca Morini,
Samantha Clarke, Alex Masters
Disruptive Media Learning Lab. (DMLL)
Coventry University
Coventry, United Kingdom
Lorenz Klopfenstein, Alessandro Bogliolo, Saverio
Delpriori
Department of Pure and Applied Sciences (DiSPeA)
University of Urbino
Urbino, Italy
Abstract—Within the field of procedural content generation
(PCG) research, the use of crowd-sensing data has, until now,
primarily been used as a means of collecting information and
generating feedback relating to player experience within games,
and game aesthetics [1], [2]. However, crowd-sensing data can
offer much more, supplying a seemingly untapped font of
information which may be used within the creation of unique
PCG game spaces or content, whilst providing a visible outlet for
the dissemination of crowd-sensed material to users. This paper
examines one such use of crowd-sensed data, the creation of a
game which will reside within the CROWD4ROADS (C4RS) [3]
application, SmartRoadSense (SRS) [4]. The authors will open
with a brief discussion of PCG. Following this, an explanation of
the features and aims of the SRS application will be provided.
Finally, the paper will introduce ‘Balance Trucks’, the SRS
game, discussing the concepts behind using crowd-sensed data
within its design, its development and use of PCG.
Keywords—game; procedural content generation; game design;
video game; racing game; crowd-sensing
I.
I
NTRODUCTION
The art of making videogames is a multidisciplinary
process, expensive in both cost and effort. A triple-A title can
require thousands of person hours to develop, and production
costs can stretch to millions of pounds. This expense gradually
increases over time, as hardware becomes more powerful,
making the requirements for developing a title ever more
complex. It therefore seems logical that developers have a
strong need to generate tools and systems which help to reduce
production costs and the amount of effort put into the creation
of videogame content [5].
Procedural content generation (PCG) within videogames
can be described as the use of a computer’s processing power
to create content via mathematical equations using small
amounts of data, or data which has been obtained indirectly
from users, as inputs. Togelius, Kastbjerg, Schedl and
Yannakakis simplify this definition, describing PCG as “the
algorithmical creation of game content with limited or indirect
user input” [6]. PCG provides several interesting benefits for
the development of commercial games. The creation of tools
which can generate content for use within a game, after some
modification by a designer or artist, has become widespread
within the industry; particularly when used with artificial
intelligence or the generation of environments [5]. PCG’s
ability to provide infinite gameplay supplies an interesting way
of increasing replayability [7]. This is seen as an important
factor in videogames as it is thought that if a title has a high
replay value, players will remain engaged and continue to play
it. These two uses of PCG reduce the effort required to
construct a videogame and thus help to lower production costs
[8]. PCG also has huge potential for generating graphical
assets [9], the authors believe that these assets could be
generated in a way which reflects the data which has been used
to drive their creation and therefore help disseminate it back to
a specific data collector.
The early use of PCG within videogames was mostly
ignored by academics, and interactions between the developers
who were utilizing it and academia were almost nonexistent.
In the last 15-20 years, academic interest in the subject has
increased dramatically [8] and researchers have addressed a
range of PCG topics. Several authors have tried to define
unified taxonomies for PCG [10], [11], and develop metaphors
which aid the understanding of design relationships within the
field [12]. Research has also been conducted into the practical
uses of content generation. These studies have examined
diverse subjects such as generating levels for platform games
[13], [14], creating missions for action adventure games [15],
generating entire worlds [16] and even attempts to build
software that can procedurally generate entire games and
gameplay [17]. Recently, research has been directed at how
gameplay levels which are able to adapt to a player’s inputs,
and thus playing style, might be created [18], [19].
Within this paper we describe how data that has been
gathered via crowd-sensing technology has been used to
generate content for use within a game included alongside the
SmartRoadSense (SRS) application [4]. To the best of our
knowledge, this represents the first time that crowd-sensing
data has been utilized as a seed for generating gameplay and
disseminating data back to users. In the following sections, we
will provide a brief overview of SRS, describing how the
application gathers and utilizes road surface data. We will then
introduce our associated game, ‘Balance Trucks’, and examine
how it uses the data collected by its parent application to
generate simple content that disseminates the data, which
individual users have gathered, back to them. Finally, we will
offer up our conclusions about the project and touch upon the
challenges it still faces as it heads towards completion.
II. S
MART
R
OAD
S
ENSE
Developed by an international team, led primarily by the
University of Urbino, Italy, as part of CROWD4ROADS
(C4RS) [3], an EU Horizon 2020 funded project,
SmartRoadSense is a crowd-sensing system for the continuous
monitoring of road quality. The application promotes active
citizenship and allows volunteers to freely participate in the
collection of valuable road infrastructure data. The system
consists of three constituent components: i) an Android/iOS
application, which collects data from the triaxial
accelerometers and GPS sensors of a smartphone and generates
location-based estimates relating to the quality of a road’s
surface; ii) a cloud-based service to which collected data is
downloaded and aggregated with existing cartographic data
from OpenStreetMap [20]; and iii) a web-based user interface,
providing access to the data (distributed under an open license,
in CSV format), and an interactive map showing a visual
representation of the data [4], [21].
The collection of SRS data (see Fig.1) begins within the
application. Once the app has been started by the user, it will
begin gathering information from the smartphones sensor’s as
the user drives. No further input is required until the end of a
journey, at which point the user can tap their device’s screen to
halt collection. The gathered data is then processed within the
smartphone to generate a single number for each second of the
journey; a “roughness index” value, which estimates the
quality of the road’s surface at given points along the recorded
track, the speed of the vehicle, and a timestamp. Each dataset is
then given a unique identifier, a track ID, and automatically
uploaded at the next available opportunity. Every 6 hours, new
tracks are aggregated with the pre-existing data on the SRS
servers and mapped using OpenStreetMap. Once this process is
complete, a new set of roughness index data for each sampled
location is published for visualization and download on the
SRS website [4], [21], [22].
Fig.1. The process of collecting SmartRoadSense data.
The application, which has been deployed for over three
years, has been continuously monitoring and collecting data
thanks to the contributions of more than 3000 users. The
system is being progressively scaled up, currently being
publicly available in several countries, including Italy, the
United Kingdom, Romania, and Greece. It has already
amassed a large volume of data, with information gathered
from over 18 million data points. At time of writing, this
provides analysis for almost 52,000 kilometers of road surface.
Within Italy and Romania, discussions with local authorities
and transport companies relating to the adoption and use of
SRS and its data have taken place, whilst in the UK, public
transport and service vehicles within Buckinghamshire have
recently begun testing the application.
III. B
ALANCE
T
RUCKS
Balance Trucks is intended to act as a tool for generating
interest in SRS, and therefore increasing its user-base. The
game will also disseminate information back to users about the
data they have personally collected. The game is primarily
aimed at children and young adults; however, it is hoped that a
broad range of ages will participate in data collection and play.
It can be accessed via SRS’s main menu and consists of a side-
scrolling racing game which utilizes the data users have
personally collected within SRS to generate playable levels.
Within single-player, the player can select one of a range of
vehicles, which are collected through gameplay, and a playable
level to race. The player will then be provided with an object
which they must transport, within their vehicle, across the level
in the fastest time possible. The game is therefore a form of
time trial in which the player competes against themselves.
Each of the games vehicles will exhibit varying statistics; some
being fast and others stable, etc. Players must therefore choose
between speed, to complete the level as quickly as possible,
versus stability, to retain the transported object. In terms of
control, players will be able to control their vehicles
acceleration, braking and, when airborne, its angle by tapping
and pressing buttons shown on screen; an on-screen display
(OSD). Throughout levels, players will be able to collect a
range of items to improve their chances of success. These
include coins, which can be used to purchase items such as
vehicles from the games offline store; boosters, which behave
in a similar manner to the power-ups in Super Mario Kart [23];
and collectible car parts, which unlock special vehicles once all
the relevant components have been discovered. All these items
will be either procedurally-generated or procedurally placed
using a mix of randomization and PCG grammars.
Once a player completes a level, they will be informed of
their finishing time and rewarded if they have beaten their
previous best. Points will be awarded based on the finishing
time, and these will be presented visually within the games
frontend (FE); the menus by which players navigate through
the game, adding a high score element into gameplay. Points
which the player scores will be added to a cumulative pot, this
total will be used to dictate the players ‘level’. The players
level will then determine the difficulty of the game and modify
the generated terrain, automatically making it easier or harder
to traverse.
Based on feedback received from case studies, multiplayer
will retain much of the gameplay evident in single-player
whilst changing the dynamic from a time trial to that of player
versus player. This change is performed by removing the need
to transport objects from the game and adding in multiplayer
systems which allow players to connect and race across terrain
generated from data on one of the connected smartphones.
Players will retain the ability to collect items within
multiplayer, and points will be awarded based upon finishing
position rather than the time to complete a level.
All the PCG elements which go towards making Balance
Trucks are generated in-app, on a player’s device. Within the
game, the largest and most important component to be
generated is the playable terrain; the surface over which players
race. The generation of this terrain utilizes two forms of PCG;
content selection, which is defined as the selection of content
from a library, in this case SRS data, and constraint
satisfaction, which uses algorithms to apply constraints to the
generated content [24]. The playable surface is generated
using 200 concurrent data entries which have been collected by
the user on their device; making each level unique to the player
and the locations they have driven. PCG methods are then
used to generate a height-map, which resembles a graph, and
modify it to remove any major spikes or troughs within the
data. Further balancing is performed by comparing each of the
data entries against the sum of its proceeding and following
neighbors divided by two. If the data’s value is greater than the
result of this sum, it is replaced with the sum itself, otherwise
the original data entry will be used. If d2 is the data to be
compared whilst d1 and d3 are its neighbors, this can be
formulated as follows:
If (d2>(d1+d3)/2, (d1+d3)/2, else d2)
To smooth terrain at the start and finish of each level, 10 blocks
of artificial data are added to the beginning and end of the
genuine SRS data. The first and final blocks of artificial data
are generated using a random number between 0.0 and 0.3.
The data in between these blocks and the SRS data is then
calculated using an equation like that used for balancing
genuine data. If b0 were to represent the first and final pieces
of artificial data, b1 the remaining artificial data to be added
and b10 the SRS data, the formula would work as follows:
b0 = random between (0.0,0.3)
b1 = (b0+b10)/2
Once the playable surface has been generated it is
populated with additional items such as coins, boosters and car
components. Each of these collectibles are procedurally-
generated and placed into a level using PCG grammars,
production rules and algorithms, which dictate the content, its
frequency and availability. To provide an example of this,
prior to balancing the game there is a 1:750 chance of a special
car component being spawned onto one of the SRS data points
used to construct the playable surface. Once a car component
spawns this chance is reduced to zero for the remainder of the
level. This simple method for spawning items also allows
balancing to be performed with ease by simply tweaking the
numbers involved.
Once the games playable surface and additional items have
been generated, the level is skinned using a series of art assets.
These have been created so that each gameplay environment
can be gradually built up in layers; foreground, playable, mid-
ground and background layers. Each of the layers are placed
slightly apart and behind the preceding layer to create a
parallaxing effect, giving an illusion of depth and movement.
The artwork used for this has been generated by team members
and students from Coventry University’s Faculty of Arts and
Humanities, under the banner of Phoenix Interactive; a
Disruptive Media Learning Lab (DMLL) project which teaches
students about the games industry by involving them in the
process of making games. The complete process for generating
gameplay within Balance Trucks and its relationship with SRS,
and its users, is shown in Fig.2.
Fig.2. The interaction between users, data, procedural content generation,
and gameplay.
IV. C
ONCLUSION
Within academic circles, there has been some discussion as
to whether content selection, which is used heavily within
Balance Trucks, offers sufficient complexity to warrant being
called PCG [24]. The authors, however, agree with Smith, who
states that “content selection, however simple, is a form of
PCG when it is used to procedurally create an environment for
the player to explore…” [25]. When used in combination with
constraint selection, this provides the fastest method for
generating content within Balance Trucks and allows SRS data
to be replicated in a visual manner within the gameplay
environment; the generated terrain closely resembling the
surface of the road from which the user collected the data. This
is one of the primary aims of the project and fulfills EU 2020
funding criteria for disseminating SRS data back to users. To
the best of our knowledge, PCG has not been used in this
context before now. Therefore, whilst this paper does not offer
up any new methods for performing PCG, we believe that the
use of crowd-sensed data to generate content using PCG
methods may offer a new source for seeding PCG elements and
generating gameplay.
Recently, work on Balance Trucks has moved into the
development stage. Testbeds have been created and these have
provided knowledge about how levels can be constructed and
led to the realization that a full testbed would be of great
benefit to the project as developments advances; creation of
such a testbed is now in progress. The design and development
of art assets for use within the games levels has also been a key
focus. As part of this process, students from Coventry
University’s were recruited and have since worked alongside
the design team to generate environment art, menu or front-end
screens and vehicles. Much work has been completed, with the
artwork for several levels, menu screens, vehicles and the on-
screen display having been delivered to the code team in
Urbino; see Fig.3 for an example of this work.
Fig.3. Balance Trucks level showing vehicle and user-interface.
Balance Trucks still faces a great many challenges. Whilst
the game design and visual style are well defined, a large
volume of art assets and coding work remains. Coding of a full
testbed is progressing and will form the foundation of the game
once complete. This will no doubt raise a new set of issues and
challenges as development progresses. Not least amongst these
challenges is the need to balance gameplay. The degree to
which this is performed, and the resulting quality of gameplay,
will be key to how the game is received. Hopefully, we will be
able to report that our efforts were successful in future
publications.
ACKNOWLEDGMENTS
The SmartRoadSense project would like to thank the
following for the creation of artwork: Kimberley Bannister,
Emilia Byrne, Nyasha Mhazo, Michael Murphy, Charlotte
Palmer, Jia Yi Tan, Vytautas Vasiliuskas, and Lina
Vysniauskaite. This project received funding from the
European Union’s Horizon 2020 research and innovation
programme under grant agreement No 687959, and the
Disruptive Media Learning Lab, UK.
REFERENCES
[1] J. Togelius, A.J. Champandard, P.L. Lanzi, M. Mateas, A. Paiva, M.
Preuss, and K.O. Stanley, “Procedural content generation: Goals,
challenges and actionable steps,” Artificial and Computational
Intelligence in Games, Dagstuhl Follow-Ups, 6, pp. 61-75, 2013.
[2] C. Pedersen, J. Togelius, and G.N. Yannakakis, “Modeling player
experience for content creation,” IEEE Transactions on Computational
Intelligence and AI in Games, 2 (1), pp. 54–67, 2010.
[3] http://www.c4rs.eu/
[4] http://smartroadsense.it/
[5] S. Oliveira, and L. Magalhães, “Adaptive content generation for games”,
Computação Gráfica e Interação (EPCGI), 2017 24º Encontro Português
de, pp. 1-8, 2017.
[6] J. Togelius, E. Kastbjerg, D. Schedl, and G.N. Yannakakis, “What is
procedural content generation? Mario on the borderline,” in PCGames
’11, Proceedings of the 2nd International Workshop on Procedural
Content Generation in Games, ACM Press, 2011.
[7] N. Barreto, A. Cardoso, and L. Roque, “Computational creativity in
procedural content generation: A state of the art survey,” In Proceedings
of the 2014 Conference of Science and Art of Video Games, 2014.
[8] J. Togelius, J. Whitehead, and R, Bidarra, “Guest editorial: procedural
content generation in games,” IEEE Transactions on Computational
Intelligence and AI in Games, 3 (3), 2011.
[9] E.J. Hastings, R.K. Guha, and K.O. Stanley, “Automatic content
generation in the galactic arms race video game,” IEEE Transactions on
Computational Intelligence and AI in Games, 1 (4), pp.245-263, 2009.
[10] J. Togelius, G.N. Yannakakis, K. O. Stanley, and C. Browne, “Search-
based procedural content generation: A taxonomy and survey,” IEEE
Transactions on Computational Intelligence and AI in Games 3 (3), pp.
172-186, 2011.
[11] M. Hendrikx, S. Meijer, J. Van Der Velden, and A. Iosup, “Procedural
content generation for games: A survey,” ACM Transactions on
Multimedia Computing, Communications, and Applications (TOMM), 9
(1), ACM Press, 2013.
[12] R. Khaled, M. J. Nelson, and P. Barr, “Design metaphors for procedural
content generation in games,” CHI '13 Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems , ACM Press, pp.
1509-1518, 2013.
[13] A.B. Moghadam, and M. K. Rafsanjani, “A genetic approach in
procedural content generation for platformer games level creation,” 2nd
Conference on Swarm Intelligence and Evolutionary Computation
(CSIEC), 2017.
[14] M. Kerssemakers, J. Tuxen, J. Togelius, G.N. Yannakakis, “A
procedural procedural level generator generator,” 2012 IEEE
Conference on Computational Intelligence and Games (CIG), 2012.
[15] J. Dormans, “Adventures in level design: Generating missions and
spaces for action adventure games,” PCGames '10 Proceedings of the
2010 Workshop on Procedural Content Generation in Games, ACM
Press, 2010.
[16] M. Nitsche, C. Ashmore, W, Hankinson, R. Fitzpatrick, J. Kelly, and K.
Margenau, “Designing procedural game spaces: A case study,” in
Proceedings of FuturePlay, 2006
[17] J. Togelius, and J. Schmidhuber, “An experiment in automatic game
design,” IEEE Symposium On Computational Intelligence and Games,
CIG '08, 2008.
[18] J. Togelius, S. Karakovskiy, J. Koutník, and J. Schmidhuber, “Super
Mario Evolution,” 2009 IEEE Symposium on Computational
Intelligence and Games, CIG2009, pp.156-161, 2009.
[19] J. Togelius, S. Karakovskiy, and R. Baumgarten, “The 2009 Mario AI
Competition,” IEEE Congress on Evolutionary Computation (CEC), pp.
1-8, 2010.
[20] http://www.openstreetmap.org/
[21] V. Freschi, S. Delpriori, L. Klopfenstein, E. Lattanzi, G. Luchetti, and
A. Bogliolo, “Geospatial data aggregation and reduction in vehicular
sensing applications: The case of road surface monitoring,” International
Conference on Connected Vehicles and Expo. ICCVE, 2014.
[22] G. Alessandroni, L. Klopfenstein, S. Delpriori, M. Dromedari, G.
Luchetti, B. Paolini, A. Seraghiti, E. Lattanzi, V. Freschi, A. Carini, and
A. Bogliolo, “Smartroadsense: Collaborative road surface condition
monitoring,” in Proceedings of the Eighth International Conference on
Mobile Ubiquitous Computing, Systems, Services and Technologies,
Ubicomm 2014. IARIA, 2014.
[23] S. Miyamoto, Super Mario Kart, Nintendo Co. Ltd, 1992.
[24] A. Sullivan, “Content Selection vs. Content Generation”, Expressive
Intelligence Studio, https://eis-blog.soe.ucsc.edu/2010/06/content-
selection-vs-content-generation/, 2010.
[25] G. Smith, “Understanding procedural content generation: A design-
centric analysis of the role of PCG in games”, Proceedings of the ACM
CHI Conference on Human Factors in Computing Systems (CHI 2014),
2014.
