Machine Learning Theory and Practice, 2026, 6(1); doi: 10.38007/ML.2026.060113.
Fan Wang
Northeastern University, Boston, 2115, MA, US
Room-scale virtual reality (VR) agents need to change their behaviour in real time according to the continuous movement of users and reduce decision-making computations within a high-refresh-rate rendering cycle. To address the problems of insufficient motion history in instantaneous observations, jitter during continuous motion, and the coupling of safety constraints and inference latency, this paper builds a low-latency closed-loop control scheme by using a gated recurrent unit (GRU) as the temporal state encoder and a continuous motion head and safety projection layer as the execution end. First, based on published real-world VR locomotion and adaptation studies, this paper shows that multi-scene temporal data are necessary for short-term trajectory prediction. Then, 3000 tests are performed on the forward link of GRU with different sequence lengths and hidden dimensions in a reference x86-64 CPU environment. The results show that the inference latency of the P95 in nine configurations is 0.129-0.515 ms; with a hidden dimension of 64 and a sequence length of 8, the P95 latency is only 0.227 ms, and thus sufficient perception, physics and rendering time can be reserved for 90 Hz and 120 Hz VR refresh rates. This paper introduces further strategies for state organisation, gating updates, action mapping, timestamp alignment and fault degradation, and provides a verifiable real-time control framework for the engineering implementation of continuous action decision-making for VR agents.
Virtual reality intelligent agent; gated loop unit; continuous action decision-making; real-time control; timing state coding
Fan Wang. Real-Time Control Implementation of Continuous Action Decision System for VR Agents Based on Gated Recurrent Units. Machine Learning Theory and Practice (2026), Vol. 6, Issue 1: 116-123. https://doi.org/10.38007/ML.2026.060113
[1] Wang L, Xu W, Liu Y, et al. Artificial Intelligence for Virtual Reality: A Review. Science China Information Sciences, 2026, 69: 111101. https://doi.org/10.1007/s11432-024-4541-9
[2] Mayor J, Calleja P, Fuentes-Hurtado F. Long Short-Term Memory Prediction of User's Locomotion in Virtual Reality. Virtual Reality, 2024, 28: 65. https://doi.org/10.1007/s10055-024-00962-9
[3] Dohan M, Mu M, Ajit S, Hill G. Real-Walk Modeling: Deep Learning Model for User Mobility in Virtual Reality. Multimedia Systems, 2024, 30: 44. https://doi.org/10.1007/s00530-023-01200-z
[4] Valmorisco S, Raya L, Sanchez A. Enabling Personalized VR Experiences: A Framework for Real-Time Adaptation and Recommendations in VR Environments. Virtual Reality, 2024, 28: 128. https://doi.org/10.1007/s10055-024-01020-0
[5] Wang, B. (2025). Methods of Load Optimization for Computer Systems Based on Physical Principles.
[6] Guo, X. (2025, March). Research on Blockchain-Based Financial AI Algorithm Integration Methods and Systems. In 2025 IEEE International Conference on Electronics, Energy Systems and Power Engineering (EESPE) (pp. 816-821). IEEE.
[7] Guo, X. (2025, April). Research on Financial Trading Algorithms Based on Deep Reinforcement Learning. In 2025 IEEE 3rd International Conference on Control, Electronics and Computer Technology (ICCECT) (pp. 1711-1715). IEEE.
[8] Jing, X. (2025, November). In-Depth Analysis of the Optimization Decision Model of Financial Investment Portfolio Based on Machine Learning Algorithms in Python Scikit-Learn Library. In International Conference on Computational Technologies for Research in Data Analytics (pp. 130-140). Cham: Springer Nature Switzerland.
[9] Liu, X. (2026). Research on the Application of Artificial Intelligence in Consumer Privacy Data Protection. Procedia Computer Science, 279, 956-965.
[10] Yuan, Y. (2026). Research on Memory Management and Dynamic Reasoning Methods for Intelligent Agents Based on Large Language Models.
[11] Yang, Y. (2026). Causal Inference-based Identification of Incremental Effects in Digital Advertising and Optimization Pathways for Resource Allocation. Advances in Computer and Communication, 7(2).
[12] Han, W. (2025, November). An Empirical Study on Building a High-precision Forecasting System for Financial Market Trends Based on the PyTorch Deep Learning Framework. In International Conference on Computational Technologies for Research in Data Analytics (pp. 518-528). Cham: Springer Nature Switzerland.