Evolution of Reactive Rules in Multi Player Computer Games Based on Imitation

Priesterjahn, Steffen; Krämer, Oliver; Weimer, Alexander; Goebels, Andreas

doi:10.1007/11539117_105

Cited by 18 publications

(7 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In this context, imitation learning is usually aimed at the development of non-player characters with more realistic and believable human-like behaviors. For instance, Priesterjahn et al [1] developed a rule-based non player character (NPC) for Quake III R through a two-step process in which a set of rules was initially derived from the data collected from human players; subsequently, the ruleset was optimized using an evolutionary algorithm. Their final result was an NPC which would behave similarly to the observed human players while also generalizing well to unseen situations.…”

Section: Related Workmentioning

confidence: 99%

“…It includes all the methods that can be employed to learn a behavior from the observation of another player such as, (i) supervised learning, which has achieved good results in several games [7], [3], (ii) neuroevolution [5], [6], (iii) rule-based evolution [1], etc. The approaches used for imitation learning can be roughly divided into direct methods [4], which apply supervised learning to extract a model from the data logged from a target behavior; indirect methods [4], which apply typical methods of computational intelligence (e.g., neuroevolution) and use the data from the target behavior to compute a similarity score used to guide the development of the imitative controller.…”

Section: Introductionmentioning

confidence: 99%

“…Imitation learning is currently regarded as a promising approach to build human-like artificial players [1], [2], [3], [4], [5], [6]. It includes all the methods that can be employed to learn a behavior from the observation of another player such as, (i) supervised learning, which has achieved good results in several games [7], [3], (ii) neuroevolution [5], [6], (iii) rule-based evolution [1], etc.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Learning drivers for TORCS through imitation using supervised methods

Cardamone

Loiacono

Lanzi

2009

2009 IEEE Symposium on Computational Intelligence and Games

View full text Add to dashboard Cite

In this paper, we apply imitation learning to develop drivers for The Open Racing Car Simulator (TORCS). Our approach can be classified as a direct method in that it applies supervised learning to learn car racing behaviors from the data collected from other drivers. In the literature, this approach is known to have led to extremely poor performance with drivers capable of completing only very small parts of a track.In this paper we show that, by using high-level information about the track ahead of the car and by predicting high-level actions, it is possible to develop drivers with performances that in some cases are only 15% lower than the performance of the fastest driver available in TORCS. Our experimental results suggest that our approach can be effective in developing drivers with good performance in non-trivial tracks using a very limited amount of data and computational resources. We analyze the driving behavior of the controllers developed using our approach and identify perceptual aliasing as one of the factors which can limit performance of our approach.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Learning drivers for TORCS through imitation using supervised methods

Cardamone

Loiacono

Lanzi

2009

2009 IEEE Symposium on Computational Intelligence and Games

View full text Add to dashboard Cite

show abstract

“…They bring up the important issue of taking into account the context and the past, in addition to the state-vector, to reduce ambiguity when deciding which action to perform. Also in the Quake game, recent research tries to improve imitation models by means of genetic algorithms (Priesterjahn, Kramer, Weimer, & Goebels, 2005).…”

Section: R Aler Et Al I Expert Systems Withmentioning

confidence: 99%

Programming Robosoccer agents by modeling human behavior

Aler

Valls

Camacho

et al. 2009

Expert Systems with Applications

View full text Add to dashboard Cite

Esta es la versión de autor de la comunicación de congreso publicada en: This is an author produced version of a paper published in: AbstractThe Robosoccer simulator is a challenging environment for artificial intelligence, where a human has to program a team of agents and introduce it into a soccer virtual environment. Most usually, Robosoccer agents are programmed by hand. In some cases, agents make use of Machine learning (ML) to adapt and predict the behavior of the opposite team, but the bulk of the agent has been preprogrammed.The main aím of this paper is to transform Robosoccer into an interactíve game and let a human control a Robosoccer agent. Then ML techniques can be used to model his/her behavior from training instances generated during the play. This model will be used later to control a Robosoccer agent, thus imitating the human behavior. We have focused our research on low-level behavior, like looking for the ball, conducting the ball towards the goal, or scoring in the presence of opponent players. Results have shown that indeed, Robosoccer agents can be controlled by programs that model human play.

show abstract

“…This is one of the reasons why developing application with different algorithms for games becomes a hot topic. Researchers have now started to introduce and implement AI algorithms in games, either by proposing new algorithms or modifying and improving existing algorithms [12,13,14,15].…”

Section: Introductionmentioning

confidence: 99%

A comparison on the performance of crossover techniques in video game

Soon

Guan

et al. 2013

2013 IEEE International Conference on Control System, Computing and Engineering

View full text Add to dashboard Cite

This paper describes the performance of four crossover operators used in evolving the required controllers in a video game. The crossover operators used in this research are the two-point crossover, the uniform crossover, the N-point crossover, and the single-point crossover. The performance of these crossover methods were tested using Infinite Mario Bros game. This video game was chosen due to the dynamicity and complexity of the game. This paper also presents a newly designed nondeterministic based Finite State Machine (FSM) method. The Mario character uses the proposed FSM as its strategy in the game. The proposed FSM is then optimized using a modified Genetic Algorithm (GA). The results showed that the required controllers were generated successfully using the proposed method. The results also showed that the N-point crossover performed well compared to the uniform crossover, the two-point crossover and the single-point crossover methods. Keywords-Artificial Intelligence (AI); Finite State Machine (FSM); Genetic Algorithm (GA); Infinite Mario Bros; Crossover Methods.I.

show abstract

Evolution of Reactive Rules in Multi Player Computer Games Based on Imitation

Cited by 18 publications

References 7 publications

Learning drivers for TORCS through imitation using supervised methods

Learning drivers for TORCS through imitation using supervised methods

Programming Robosoccer agents by modeling human behavior

A comparison on the performance of crossover techniques in video game

Contact Info

Product

Resources

About