Evolving Interesting Maps for a First Person Shooter

1. Evolving Interesting Maps
for a First Person Shooter
Luigi Cardamone, Georgios Yannakakis,
Julian Togelius and Pier Luca Lanzi
EVO*-2011

2. Motivations 2
 Game content?
 In many games, the availability of a big quantity of content
is crucial
 Unfortunately, the creation of the game content is a very
time-consuming activity
Several game designers are required
Some content need to be tested
Luigi Cardamone et al. – EVO* 2011

3. SBPCG 3
 Search-based Procedural Content Generation (SBPCG):
PCG (algorithms procedures for generating content)
+
Search algorithms (e.g. genetic algorithms)
 SBPCG can be a valid tool:
To support the game designers
To extends the game with unlimited content
 Challenges:
How to represent the content?
Which is the best representation to be evolved with
genetic algorithms?
How to evaluate the game content?
Luigi Cardamone et al. – EVO* 2011

4. This work 4
 Focus on a First Person Shooter
 We Apply SBPCG to evolve maps that can be interesting for
the players
 We focus on maps:
 For Multi-player deathmatch
 Single floor
 We introduce 4 genotype representations for the maps
 We propose a fitness to measure the interestingness of a
given map
Luigi Cardamone et al. – EVO* 2011

5. Cube-2 6
 Sophisticated First Person
Shooter Game:
 Deathmatch
 Team Deathmatch
 Capture The Flag
 Fast Engine allowing for
in-game geometry editing
 Open-source
Luigi Cardamone et al. – EVO* 2011

6. Problem Setting 7
 Map elements:
Structural elements:
• Rooms Part of the
• Corridors representation
• Elevators/stairs
Game Elements:
• Weapons Algorithmically
• Bonus generated
• Spawn-points
Graphical Elements:
• Non-interactive objects
• Lights
• Textures
Luigi Cardamone et al. – EVO* 2011

7. Representations 8
 Phenotype:
Single floor map (2D)
A big matrix of blocks that can
form corridors and rooms
 4 Genotype Representations:
All-White
All-Black
Grid
Random Digger
 Mapping procedures are necessary to convert the genotype
to the phenotype:
Check unreachable parts
Place uniformly weapons, bonus, and spawn points
Luigi Cardamone et al. – EVO* 2011

8. All-White 9
 The initial space is free
 The genome encodes a series of obstacles that are added to
the map
 Each obstacle is a wall represented as:
x,y (start position)
Length (positive for horizontal, negative for vertical)
<(10,10,200),(90,90,-100),…>
Luigi Cardamone et al. – EVO* 2011

9. All-Black
 The initial space is all black (full of obstacles)
 The genome encodes a series of free spaces that are added
to the map
 Each free space can be:
 A square arena (x,y,size)
 A rectangular corridor (x,y,length)
Luigi Cardamone et al. – EVO* 2011

10. Grid
 The initial space has a grid of walls that
creates NxN squares
 The genome specifies for each wall of the
grid if it should be ON or OFF
110111
110111
111011
101111
111111
111111
Luigi Cardamone et al. – EVO* 2011

11. Random Digger
 The initial space is all black (full of obstacles)
 An agent starts moving randomly and “dig”
every place visited
 The genome encodes the policy which
controls this agent
Luigi Cardamone et al. – EVO* 2011

12. Removing Unreachable parts 13
 All the first 3 representations may lead to maps with
unreachable parts:
 Remove the parts that are not reachable:
1. Compute the connected spaces
2. Find the biggest space
3. Remove the zones that are not reachable from the main
one
1
3
2
Luigi Cardamone et al. – EVO* 2011

13. Fitness 14
 Based on the Fighting Time
 Also the size of the map is taken into account
Fitness = FightTime + Size
 To evaluate the FightTime:
Simulated match of 10 minutes
4 bots (need navigation points)
spawn start-fight killed spawn start-fight killed
Fighting Time Fighting Time
Luigi Cardamone et al. – EVO* 2011

14. Results

15. 16
Best Maps evolved with representation All-White
Best Maps evolved with representation All-Black
Luigi Cardamone et al. – EVO* 2011

16. 17
Best Maps evolved with representation Grid
Best Maps evolved with representation RandomDigger
Luigi Cardamone et al. – EVO* 2011

17. Comparison 18
 10 runs for each representation
 Grid and All-White are the representations which achieve the
highest fitness
> >
Luigi Cardamone et al. – EVO* 2011

18. Maps Loaded in Cube 2 19
Luigi Cardamone et al. – EVO* 2011

19. Game-play example 20
Luigi Cardamone et al. – EVO* 2011

20. Conclusions 21
 All the representations are able to evolve playable maps
 Preliminary user testing suggests that the fitness push
towards interesting maps
 All-White is the representation that is able to evolve the best
maps
 Future works:
Use additional fitness functions
Extends the analysis using more human players
Luigi Cardamone et al. – EVO* 2011

21. Thank you! Questions?