TOAD-GAN: A Flexible Framework for Few-Shot Level Generation in Token-Based Games

Schubert, Frederik; Awiszus, Maren; Rosenhahn, Bodo

doi:10.1109/tg.2021.3069833

Cited by 8 publications

(7 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Again, the agent is trained for 5 different random seeds and with 100 training instances. In CARLMarioEnv different instances (Mario levels) are created by using TOAD-GAN [50]. By varying the noise input vector for TOAD-GAN we can generate different levels and the greater the noise, the greater the differences to the original level.…”

Section: Additional Experimental Resultsmentioning

confidence: 99%

“…In order to gain insight on how the context and its augmentation influences the agent's learning and behavior, we provide several benchmarks in CARL. As first benchmarks we include and contextually extend classic control and box2d environments from OpenAI Gym [7], Google Brax' walkers [17], a RNA folding environment [48] as well as Super Mario levels [3,50]. See Figure 2 for an overview of included environments.…”

Section: The Carl Benchmarksmentioning

confidence: 99%

“…Besides physical simulation environments, CARL provides two more specific, challenging environments. The first is the CARLMarioEnv environment built on top of the TOAD-GAN level generator [3,50]. It provides a procedurally generated game playing environment (similarly to the ones discussed in Section 4) that allows customization of the generation process.…”

Section: A Benchmark Categoriesmentioning

confidence: 99%

“…(c) Brax [17] Ant and HalfCheetah, Fetch and Humanoid, Grasp and UR5E. (d) Super Mario[3,50] and RNADesign[48].…”

mentioning

confidence: 99%

See 3 more Smart Citations

CARL: A Benchmark for Contextual and Adaptive Reinforcement Learning

Benjamins¹,

Eimer²,

Schubert³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

While Reinforcement Learning has made great strides towards solving ever more complicated tasks, many algorithms are still brittle to even slight changes in their environment. This is a limiting factor for real-world applications of RL. Although the research community continuously aims at improving both robustness and generalization of RL algorithms, unfortunately it still lacks an open-source set of well-defined benchmark problems based on a consistent theoretical framework, which allows comparing different approaches in a fair, reliable and reproducible way. To fill this gap, we propose CARL, a collection of well-known RL environments extended to contextual RL problems to study generalization. We show the urgent need of such benchmarks by demonstrating that even simple toy environments become challenging for commonly used approaches if different contextual instances of this task have to be considered. Furthermore, CARL allows us to provide first evidence that disentangling representation learning of the states from the policy learning with the context facilitates better generalization. By providing variations of diverse benchmarks from classic control, physical simulations, games and a real-world application of RNA design, CARL will allow the community to derive many more such insights on a solid empirical foundation.

show abstract

Section: Additional Experimental Resultsmentioning

confidence: 99%

Section: The Carl Benchmarksmentioning

confidence: 99%

Section: A Benchmark Categoriesmentioning

confidence: 99%

“…(c) Brax [17] Ant and HalfCheetah, Fetch and Humanoid, Grasp and UR5E. (d) Super Mario[3,50] and RNADesign[48].…”

mentioning

confidence: 99%

See 2 more Smart Citations

CARL: A Benchmark for Contextual and Adaptive Reinforcement Learning

Benjamins¹,

Eimer²,

Schubert³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Overall, future work needs to focus on building tools that enable a wider variety of users to control such models (e.g., enabling the use of UI elements and interfaces rather than defining one-hot encoded labels and coding up fitness functions for evolution). Such tools could take inspiration from the large body of existing co-creative systems [80], particularly those based on ML, such as Morai Maker [81] and TOAD-GUI [82], both for making Mario levels, the tool from Schrum et al [83] for interactively exploring the latent spaces of GANs trained on Mario and Zelda, as well as Lode Encoder [84], a VAEbased tool for generating Lode Runner [85] levels. Note that in all these examples, the tools only work in one game/domain and thus cannot be considered as tools for PCG-KT.…”

Section: Design Tools and Controllabilitymentioning

confidence: 99%