Action Spaces in Deep Reinforcement Learning to Mimic Human Input Devices

Pleines, Marco; Zimmer, Fränk; Berges, Vincent-Pierre

doi:10.1109/cig.2019.8848080

Cited by 4 publications

(3 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Jednak nie zostały one szczegółowo opisane, ani nie wskazano zastosowanych algorytmów uczenia. Platforma Unity oraz biblioteka Unity ML-Agents Toolkit posłużyła również do badania skuteczności sposobów definiowania przestrzeni akcji agenta [9]. Wśród narzędzi pozwalających na testowanie algorytmów uczenia ze wzmocnieniem w symulowanych środowiskach, obok Unity, można wyróżnić platformę Arcade Learning Environment [7,[10][11][12][13] oraz OpenAI Gym [14].…”

Section: Wstępunclassified

“…Akcje agenta zaimplementowano jako dyskretne, stosując gałęzie akcji (z ang. action branching), które pozwalają uprościć przestrzeń akcji, rozdzielając ją na grupy, z których agent może wybierać wartości jednocześnie [9,15]. Ponadto akcje podejmowane są co 5 kroków środowiska i są one automatycznie powtarzane pomiędzy kolejnymi decyzjami.…”

Section: Projekt śRodowiskaunclassified

See 1 more Smart Citation

Analysis of the possibilities for using machine learning algorithms in the Unity environment

Litwynenko¹,

Plechawska–Wójcik

2021

jcsi

View full text Add to dashboard Cite

Reinforcement learning algorithms are gaining popularity, and their advancement is made possible by the presence of tools to evaluate them. This paper concerns the applicability of machine learning algorithms on the Unity platform using the Unity ML-Agents Toolkit library. The purpose of the study was to compare two algorithms: Proximal Policy Optimization and Soft Actor-Critic. The possibility of improving the learning results by combining these algorithms with Generative Adversarial Imitation Learning was also verified. The results of the study showed that the PPO algorithm can perform better in uncomplicated environments with non-immediate rewards, while the additional use of GAIL can improve learning performance.

show abstract

Section: Wstępunclassified

Section: Projekt śRodowiskaunclassified

Analysis of the possibilities for using machine learning algorithms in the Unity environment

Litwynenko¹,

Plechawska–Wójcik

2021

jcsi

View full text Add to dashboard Cite

show abstract

“…As stated in the introduction, Jacob et al [3] argued that industry does not need agents build to "beat the game", but rather to produce credible and human-like behaviors. Results such as [11]- [13] are other notable examples of applying DRL to commercial video games, as well as the DeepCrawl work [14], [15], a game environment specicially designed for studying the applicability of DRL in video game production. Within this context, Procedural Content Generation (PCG) has recently gained a lot of attention: works like [16]- [18] have noticed that diverse environment distributions are essential to adequately train and evaluate RL agents for video game production, as these kinds of environments enable generalization of agents when faced with design changes [15].…”

Section: Related Workmentioning

confidence: 99%

Policy Fusion for Adaptive and Customizable Reinforcement Learning Agents

Sestini¹,

Kuhnle²,

Bagdanov³

2021

Preprint

View full text Add to dashboard Cite

In this article we study the problem of training intelligent agents using Reinforcement Learning for the purpose of game development. Unlike systems built to replace human players and to achieve super-human performance, our agents aim to produce meaningful interactions with the player, and at the same time demonstrate behavioral traits as desired by game designers. We show how to combine distinct behavioral policies to obtain a meaningful "fusion" policy which comprises all these behaviors. To this end, we propose four different policy fusion methods for combining pre-trained policies. We further demonstrate how these methods can be used in combination with Inverse Reinforcement Learning in order to create intelligent agents with specific behavioral styles as chosen by game designers, without having to define many and possibly poorly-designed reward functions. Experiments on two different environments indicate that entropy-weighted policy fusion significantly outperforms all others. We provide several practical examples and use-cases for how these methods are indeed useful for video game production and designers.

show abstract