Article: Application of deep reinforcement learning in intelligent interaction design of virtual practice scenarios for labour education in colleges and universities Journal: International Journal of Continuing Engineering Education and Life-Long Learning (IJCEELL) 2026 Vol.36 No.9 pp.240 - 263 Abstract: This paper proposes an adaptive interaction strategy model based on deep Q-network (DQN) to break through the limitations of traditional static design. The proposed model first extracts behavioural features from students' operation data through convolutional neural network (CNN) to construct state representation. Then, a reward function with task completion and learning efficiency as the goals is designed. Next, the deep Q-network algorithm is used to train the intelligent agent to dynamically adjust the task difficulty, environmental variables, and guidance strategies. Finally, the model is deployed to the virtual platform to achieve real-time optimisation interaction. Experimental results show that the proposed model extracts students' operation features through CNN, and dynamically adjusts task difficulty and guidance feedback based on DQN, so that students' participation score reaches 92 points, and the average learning efficiency score reaches 97 points, providing an efficient and personalised solution for labour education in colleges and universities. Inderscience Publishers - linking academia, business and industry through research

Title: Application of deep reinforcement learning in intelligent interaction design of virtual practice scenarios for labour education in colleges and universities

Authors: Wei Wang; Rui Liu; Zhuo Liu

Addresses: Institute of Physical Education, Xinjiang Normal University, Urumqi – 830054, Xinjiang, China ' Collage of Foreign Language, Xinjiang Normal University, Urumqi – 830017, Xinjiang, China ' College of Fine Arts, Xinjiang Normal University, Urumqi – 830054, Xinjiang, China

Abstract: This paper proposes an adaptive interaction strategy model based on deep Q-network (DQN) to break through the limitations of traditional static design. The proposed model first extracts behavioural features from students' operation data through convolutional neural network (CNN) to construct state representation. Then, a reward function with task completion and learning efficiency as the goals is designed. Next, the deep Q-network algorithm is used to train the intelligent agent to dynamically adjust the task difficulty, environmental variables, and guidance strategies. Finally, the model is deployed to the virtual platform to achieve real-time optimisation interaction. Experimental results show that the proposed model extracts students' operation features through CNN, and dynamically adjusts task difficulty and guidance feedback based on DQN, so that students' participation score reaches 92 points, and the average learning efficiency score reaches 97 points, providing an efficient and personalised solution for labour education in colleges and universities.

Keywords: deep reinforcement learning; adaptive interaction model; labour education; virtual practice; student participation.

DOI: 10.1504/IJCEELL.2026.153603

International Journal of Continuing Engineering Education and Life-Long Learning, 2026 Vol.36 No.9, pp.240 - 263

Received: 18 Jul 2025
Accepted: 29 Jan 2026
Published online: 18 May 2026 *

Title: Application of deep reinforcement learning in intelligent interaction design of virtual practice scenarios for labour education in colleges and universities

Keep up-to-date