Article: Behaviour control of VR virtual characters via path planning and reinforcement learning Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.22 pp.71 - 88 Abstract: Compelling virtual reality experiences depend on characters that move and interact in a natural, socially aware manner. In crowded indoor environments, these characters must navigate efficiently while adhering to unspoken social norms - maintaining personal space, yielding at doorways, and anticipating human movement - a dual requirement that current approaches fail to fulfil simultaneously. This paper presents a framework that integrates a neural distance field into a hierarchical reinforcement learning policy, thereby jointly reasoning about spatial traversability and generating socially compliant motions. A cognitive-load-theory-based attention gating mechanism constrains the character to focus only on the most interaction-relevant individuals, preventing information overload. Evaluated on the THÖR-MAGNI dataset of real human-robot trajectories, our method achieves 91.3% task success, reduces path completion time by 23.6%, boosts passing etiquette score by 42.6%, and lowers collisions by 40.2% over the strongest baseline. This fills a crucial gap in virtual character control. Inderscience Publishers - linking academia, business and industry through research

Title: Behaviour control of VR virtual characters via path planning and reinforcement learning

Authors: Junli Wang

Addresses: School of Digital Arts, Wuxi Vocational College of Science and Technology, Wuxi, 214000, China

Abstract: Compelling virtual reality experiences depend on characters that move and interact in a natural, socially aware manner. In crowded indoor environments, these characters must navigate efficiently while adhering to unspoken social norms - maintaining personal space, yielding at doorways, and anticipating human movement - a dual requirement that current approaches fail to fulfil simultaneously. This paper presents a framework that integrates a neural distance field into a hierarchical reinforcement learning policy, thereby jointly reasoning about spatial traversability and generating socially compliant motions. A cognitive-load-theory-based attention gating mechanism constrains the character to focus only on the most interaction-relevant individuals, preventing information overload. Evaluated on the THÖR-MAGNI dataset of real human-robot trajectories, our method achieves 91.3% task success, reduces path completion time by 23.6%, boosts passing etiquette score by 42.6%, and lowers collisions by 40.2% over the strongest baseline. This fills a crucial gap in virtual character control.

Keywords: virtual character control; hierarchical reinforcement learning; HRL; neural distance field; NDF; social navigation; cognitive attention.

DOI: 10.1504/IJRIS.2026.155993

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.22, pp.71 - 88

Received: 19 May 2026
Accepted: 22 Jun 2026
Published online: 28 Aug 2026 *

Title: Behaviour control of VR virtual characters via path planning and reinforcement learning

Keep up-to-date