Injecting human knowledge is an effective way to accelerate reinforcement learning (RL). However, these methods are underexplored. This article presents our discovery that abstract forward model [thought-game (TG)] combined with transfer way. We take StarCraft II as study environment. With the help of a designed TG, agent can learn 99% win-rate on 64×64 map against Level-7 ...

Applying game-tree search techniques to RTS games poses a significant challenge, given the large branching factors involved. This paper studies an approach to incorporate knowledge learned offline from game replays to guide the search process. Specifically, we propose to learn Naive Bayesian models predicting the probability of action execution in different game states, and use them to inform t...

In this paper, we discuss a part of participatory culture that so far has not received much attention in the academic world; it is the writing and reading of game fan fiction. The focus in this paper is on fan fiction, based on three different games that represent three different game genres: Tetris, StarCraft and Dreamfall: The Longest Journey. The aim is to advance our understanding of how pl...

It has been well recognized that human makes use of both declarative memory and procedural memory for decision making and problem solving. In this paper, we propose a computational model with the overall architecture and individual processes for realizing the interaction between the declarative and procedural memory based on self-organizing neural networks. We formalize two major types of memor...

A number of applications have emerged over recent years that use datagram transport. These applications include real time video conferencing, Internet telephony, and online games such as Quake and StarCraft. These applications are all delay sensitive and use unreliable datagram transport. Applications that are based on reliable transport can be secured using TLS, but no compelling alternative e...

The exploitation of extra state information has been an active research area in multi-agent reinforcement learning (MARL). QMIX represents the joint action-value using a non-negative function approximator and achieves best performance on StarCraft II micromanagement testbed, common MARL benchmark. However, our experiments demonstrate that, some cases, performs sub-optimally with A2C framework, ...

From an AI point of view, Real-Time Strategy (RTS) games are hard because they have enormous state spaces, they are real-time and partially observable. In this paper, we explore an approach to deploy gametree search in RTS games by using game state abstraction, and explore the effect of using different abstractions over the game state. Different abstractions capture different parts of the game ...

From an AI point of view, Real-Time Strategy (RTS) games are hard because they have enormous state spaces, they are real-time and partially observable. In this paper, we present an approach to deploy gametree search in RTS games by using game state abstraction. We propose a high-level abstract representation of the game state, that significantly reduces the branching factor when used for game-t...

The task of keyhole (unobtrusive) plan recognition is central to adaptive game AI. “Tech trees” or “build trees” are the core of real-time strategy (RTS) game strategic (long term) planning. This paper presents a generic and simple Bayesian model for RTS build tree prediction from noisy observations, which parameters are learned from replays (game logs). This unsupervised machine learning appro...