Bootstrapping Conditional GANs for Video Game Level Generation

Abstract

Generative Adversarial Networks (GANs) have shown impressive results for image generation. However, GANs face challenges in generating contents with certain types of constraints, such as game levels. Specifically, it is difficult to generate levels that have aesthetic appeal and are playable at the same time. Additionally, because training data usually is limited, it is challenging to generate unique levels with current GANs. In this paper, we propose a new GAN architecture named Conditional Embedding Self-Attention Generative Adversarial Network (CESAGAN) and a new bootstrapping training procedure. The CESAGAN is a modification of the self-attention GAN that incorporates an embedding feature vector input to condition the training of the discriminator and generator. This allows the network to model non-local dependency between game objects, and to count objects. Additionally, to reduce the number of levels necessary to train the GAN, we propose a bootstrapping mechanism in which playable generated levels are added to the training set. The results demonstrate that the new approach does not only generate a larger number of levels that are playable but also generates fewer duplicate levels compared to a standard GAN.

Level generation for angry birds with sequential VAE and latent variable evolution

“…The existing level generation techniques that employ DGMs have been applied to tile-based game domains wherein the levels are clearly divided into meshes and all blocks have the same size or the same unit size, and the blocks can be set by specifying the index of the level array. Therefore, the existing approaches such as [25][26][27] encode the tile-based level as an image and successfully use DGMs for image processing. However, the levels of Angry Birds are specified by real numbers, and a large number of meshes are needed to represent the exact stage.…”

“…PCGML includes approaches that utilize n-gram [4] with a graphical probability model [9], generative adversarial networks (GAN) [7,26,27], and variational autoencoders (VAEs) [25]. Volz et al [27] applied the Wasserstein GAN (WGAN) to generate levels for Super Mario Bros (SMB).…”

Procedural Level Generation for Sokoban via Deep Learning: An Experimental Study

“…Given the scarcity of level datasets, the bootstrapping method [4] scavenge the model output for playable levels to be used for training in the upcoming iterations. In addition, conditional embeddings [4] were proposed to improve the generator's quality. To train without levels, Generative Playing Networks (GPN) [7] learns to generate levels using feedback from an agent trained to play the generated levels.…”

“…The benchmark will be Sokoban level generation at size 7 × 7. The methods will be bootstrapping conditional generative adversarial networks [4], controllable & non-controllable procedural content generation via reinforcement learning (PCGRL) [5], [6] and generative playing networks (GPN) [7]. The bootstrapping method will be used with Variational Autoencoders (VAE) [8], Generative Adversarial Networks (GAN) [9] and VAEGANs [10] (which combines VAEs and GANs).…”

“…1) Adapt the bootstrapping method [4] and Generative Playing Networks [7] to Sokoban level generation. 2) Propose Diversity Sampling to improve the generators ability to generate levels with diverse solutions.…”