Article: Dynamic resource allocation in smart laboratories based on multi-agent reinforcement learning Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.60 pp.90 - 113 Abstract: To address issues such as uneven equipment utilisation and delayed user demand response in traditional smart laboratory resource allocation, this paper proposes a multi-agent deep reinforcement learning framework based on attention mechanisms. The research motivation stems from the collaborative scheduling challenges posed by heterogeneous equipment and dynamic task requests. This method employs a centralised training and distributed execution architecture, enabling agents to learn cooperative strategies in partially observable environments. It further incorporates a demand forecasting module to enhance allocation foresight. Experiments on public datasets and simulation environments demonstrate that the proposed method significantly outperforms traditional genetic algorithms and single-agent reinforcement learning approaches in both resource allocation quality (normalised discounted cumulative gain @5 reached 0.87) and overall utilisation (area under the curve improvement of 15.2%), validating its effectiveness and adaptability in complex laboratory scenarios. Inderscience Publishers - linking academia, business and industry through research

Title: Dynamic resource allocation in smart laboratories based on multi-agent reinforcement learning

Authors: Hua Yang; Xiangdong Liang; Jingwei Li

Addresses: Shanxi Police College, Taiyuan, 030001, China ' Shanxi Police College, Taiyuan, 030001, China ' Shanxi Police College, Taiyuan, 030001, China

Abstract: To address issues such as uneven equipment utilisation and delayed user demand response in traditional smart laboratory resource allocation, this paper proposes a multi-agent deep reinforcement learning framework based on attention mechanisms. The research motivation stems from the collaborative scheduling challenges posed by heterogeneous equipment and dynamic task requests. This method employs a centralised training and distributed execution architecture, enabling agents to learn cooperative strategies in partially observable environments. It further incorporates a demand forecasting module to enhance allocation foresight. Experiments on public datasets and simulation environments demonstrate that the proposed method significantly outperforms traditional genetic algorithms and single-agent reinforcement learning approaches in both resource allocation quality (normalised discounted cumulative gain @5 reached 0.87) and overall utilisation (area under the curve improvement of 15.2%), validating its effectiveness and adaptability in complex laboratory scenarios.

Keywords: multi-agent reinforcement learning; MARL; smart laboratory; dynamic resource allocation; attention mechanism.

DOI: 10.1504/IJICT.2026.153934

International Journal of Information and Communication Technology, 2026 Vol.27 No.60, pp.90 - 113

Received: 09 Feb 2026
Accepted: 14 Mar 2026
Published online: 08 Jun 2026 *

Title: Dynamic resource allocation in smart laboratories based on multi-agent reinforcement learning

Keep up-to-date