Speakers:

Hiroyuki Yabe (Future Technology Group, Game Service R&D)

Yutaro Miyauchi (Future Technology Group, Game Service R&D)

Hiroki Nakahara (Foundational Systems and Experiences Engineering Division, S Section, Department 3)

Presentation Overview

At the recent CEDEC 2024 (Computer Entertainment Developers Conference 2024), we presented AI technology that autonomous gameplay on PlayStation®5 (PS5®), replicating human-player conditions. CEDEC, one of Japan’s largest conferences for game developers and entertainment industry professionals, was held this year at Pacifico Yokohama from August 21-23, 2024.

Presentation Details

PS5 features numerous functionalities, and verifying the proper operation of each requires considerable effort and cost. To address these challenges, we have focused on automating testing processes to lower costs and identify defects earlier in development. In addition to traditional automation methods, we have recently begun exploring the use of cutting-edge AI technologies, such as deep learning, to automate Gameplay.

This presentation focused on the AI technologies we have developed for automating system software function tests on PS5. These tests must be conducted in an environment identical to that of end users. Thus, our automation system was designed to replicate gameplay using only the information available to ordinary users.

We developed an autonomous gameplay system that exclusively utilizes on-screen information. This system combines various technical components, including AI model development through imitation learning, recording and replaying controller operation histories, and model switching using image recognition. We also showcased videos demonstrating autonomous gameplay on actual game titles and shared specific insights gained through imitation learning.

With the increasing scale and complexity of games, the cost of quality assurance testing has become a growing concern across the gaming industry. We hope the knowledge gained through our automation technology will contribute to more efficient quality assurance testing and cost reduction.

Presentation Highlights

The presentation was well-received, with numerous questions from industry professionals during the Q&A session.

In particular, there was significant interest in the application of AI technologies, highlighting the industry’s keen focus on test automation using AI. Feedback emphasized the uniqueness of our research in leveraging classical learning methods to achieve practical performance, which many found highly valuable.

We aim to incorporate the insights and feedback gained from this presentation to further advance our technology.

Presentation Materials

Presentation materials are available on CEDiL (login required) (Japanese only)