Deep Reinforcement Learning for Procedural Content Generation of 3D Virtual Environments

Abstract

This work presents a deep reinforcement learning (DRL) approach for procedural content generation (PCG) to automatically generate three-dimensional (3D) virtual environments that users can interact with. The primary objective of PCG methods is to algorithmically generate new content in order to improve user experience. Researchers have started exploring the use of machine learning (ML) methods to generate content. However, these approaches frequently implement supervised ML algorithms that require initial datasets to train their generative models. In contrast, RL algorithms do not require training data to be collected a priori since they take advantage of simulation to train their models. Considering the advantages of RL algorithms, this work presents a method that generates new 3D virtual environments by training an RL agent using a 3D simulation platform. This work extends the authors’ previous work and presents the results of a case study that supports the capability of the proposed method to generate new 3D virtual environments. The ability to automatically generate new content has the potential to maintain users’ engagement in a wide variety of applications such as virtual reality applications for education and training, and engineering conceptual design.

Current Aspects of Using Artificial Intelligence in Digital Games and Computer Graphics Content Control

“…As a matter a fact, using deep reinforcement learning in playing digital games has one crucial flaw -if the environment that agent trained on is not exactly the same as in the actual game -the method overfits. By using procedural content creation in their training, agent performance became significantly more general [16].…”

The Influence of E-Scaffolding Sources in a Mobile Learning Environment on Students’ Design Skills and the Technology Fatigue Associated with a 3D Virtual Environment

“…Learners have the ability to experience real environments and be actively involved with the components of the virtual environment, thereby increasing their immersion, interaction, and sense of realism [43]. Three-dimensional virtual environments are an effective way to create interactive environments in which abstract concepts and skills can be developed through problem solving, exploration, and experimentation [44][45][46]. Tilhou et al [47] defined 3D virtual environments as 3D interactive spaces that can be used for various purposes.…”

A reinforcement learning approach to adaptive remediation in online training

“…Artificial intelligence (AI) will play a central role in modernized training systems for the military. The transformative potential of AI-enhanced training and education is made possible by advancements in a wide array of capabilities, including natural language processing, 1 multimodal learning analytics, 2 automated scenario generation, 3 and adaptive decision-making, 4 that can be leveraged to create engaging and effective training experiences that are tailored to individual learners and teams. Numerous studies have shown that when designed effectively, adaptive instructional systems (AISs), such as intelligent tutoring systems, can be as effective as a skilled human tutor at supporting student learning.…”