Probabilistic opponent-model search

Donkers, H.H.L.M.

doi:10.1016/s0020-0255(01)00133-5

Cited by 23 publications

(17 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Most previous techniques focused on a particular model of interaction, such as repeated games [9-11, 13, 31] two-player, zero-sum, perfect information games [6,12,21,23,37,41] imperfect information games [4] , and market systems [40,47]. Stone et al [44] and Bruce et al [5] applied Opponent Modelling techniques to a robotic soccer game, which is a dynamic environment with continuous action space and incomplete information.…”

Section: Related Workmentioning

confidence: 99%

“…Opponent models are represented as probabilistic automata in [13]. Some approaches utilize a mixed model of the opponent, representing uncertainty regarding the model by maintaining a probability distribution over possible models [11,12]. Finally, several techniques work with recursive models of the opponent, in which the model also includes the model the opponent holds about the agent, and so on recursively [7,17,48].…”

Section: Related Workmentioning

confidence: 99%

“…The more common approach uses the model to simulate the actual opponent actions, commonly in lookahead search. This simulation can be deterministic [15,16,21,22,44], probabilistic [4,11,12,37,41], or recursive [7,8,17,48].…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Learning and Exploiting Relative Weaknesses of Opponent Agents

Markovitch

Reger

2005

Auton Agent Multi-Agent Syst

View full text Add to dashboard Cite

Agents in a competitive interaction can greatly benefit from adapting to a particular adversary, rather than using the same general strategy against all opponents. One method of such adaptation is Opponent Modeling, in which a model of an opponent is acquired and utilized as part of the agent's decision procedure in future interactions with this opponent. However, acquiring an accurate model of a complex opponent strategy may be computationally infeasible. In addition, if the learned model is not accurate, then using it to predict the opponent's actions may potentially harm the agent's strategy rather than improving it. We thus define the concept of opponent weakness, and present a method for learning a model of this simpler concept. We analyze examples of past behavior of an opponent in a particular domain, judging its actions using a trusted judge. We then infer a weakness model based on the opponent's actions relative to the domain state, and incorporate this model into our agent's decision procedure. We also make use of a similar self-weakness model, allowing the agent to prefer states in which the opponent is weak and our agent strong; where we have a relative advantage over the opponent. Experimental results spanning two different test domains demonstrate the agents' improved performance when making use of the weakness models.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Learning and Exploiting Relative Weaknesses of Opponent Agents

Markovitch

Reger

2005

Auton Agent Multi-Agent Syst

View full text Add to dashboard Cite

show abstract

“…Second, ambient games have been shown to benefit from player behavioural analysis for adapting the game context [92,24]. Third, in classic games (e.g., Chess), it has been shown that accurately modelling the opponent player can increase the playing strength [28,29], but moreover can be applied for scaling the playing strength to be appropriate to the human player [116], for entertainment purposes.…”

Section: Generalisation To Other Domainsmentioning

confidence: 99%

Personalised gaming

Bakkes

Tan

Pisan

2012

Proceedings of the 8th Australasian Conference on Interactive Entertainment: Playing the System

View full text Add to dashboard Cite

This article focuses on personalised games, which we define as games that utilise player models for the purpose of tailoring the game experience to the individual player. The main contribution of the article is a motivation for personalised gaming, supported by an extensive overview of scientific literature. The motivatin concerns (a) the psychological foundation, (b) the effect on player satisfaction, (c) the contribution to game development, and (d) the requirement for achieving ambitions. The provided overview of scientific literature goes into the subject of player modelling, as well as eight adaptive components: (1) space adaptation, (2) mission / task adaptation, (3) character adaptation, (4) game mechanics adaptation, (5) narrative adaptation, (6) music / sound adaptation, (7) player matching (multiplayer), and (8) difficulty scaling. In the concluding sections, the relationship to procedural content generation is discussed, as well as the generalisation to other domains.

show abstract

“…PIDM is based on the simulation approach, where models of other agents' strategies are used to simulate their decision making in order to predict their actions (Iida et al, 1993;Iida et al, 1994;Donkers et al, 2001;Donkers, 2003;Carmel & Markovitch, 1998;Carmel and Markovitch, 1996a). PIDM extends the previous works by allowing any combination of opp-agents and co-agents and by combining search with model-based Monte Carlo sampling for partially observable states.…”

Section: Bridge Biddingmentioning

confidence: 99%

Learning to bid in bridge

Amit

Markovitch

2006

Mach Learn

View full text Add to dashboard Cite

Bridge bidding is considered to be one of the most difficult problems for gameplaying programs. It involves four agents rather than two, including a cooperative agent. In addition, the partial observability of the game makes it impossible to predict the outcome of each action. In this paper we present a new decision-making algorithm that is capable of overcoming these problems. The algorithm allows models to be used for both opponent agents and partners, while utilizing a novel model-based Monte Carlo sampling method to overcome the problem of hidden information. The paper also presents a learning framework that uses the above decision-making algorithm for co-training of partners. The agents refine their selection strategies during training and continuously exchange their refined strategies. The refinement is based on inductive learning applied to examples accumulated for classes of states with conflicting actions. The algorithm was empirically evaluated on a set of bridge deals. The pair of agents that co-trained significantly improved their bidding performance to a level surpassing that of the current state-of-the-art bidding algorithm.

show abstract

Probabilistic opponent-model search

Cited by 23 publications

References 13 publications

Learning and Exploiting Relative Weaknesses of Opponent Agents

Learning and Exploiting Relative Weaknesses of Opponent Agents

Personalised gaming

Learning to bid in bridge

Contact Info

Product

Resources

About