A simulated reinforcement learning framework combining generative modeling with deep policy gradients has outperformed conventional transmission strategies in maximizing the energy efficiency of dynamic indoor visible light communication systems. In research published in the Journal on Advances in Signal Processing, authors Bohao Tang, Yi Yu, Wenwen Xie, and Weicheng Chu evaluated an improved deep deterministic policy gradient framework applied to ceiling-mounted light-emitting diodes that simultaneously illuminate indoor spaces and transmit data to multiple mobile terrestrial users. According to the reported simulation results, the proposed computational method achieved superior energy efficiency when compared directly against two conventional baseline strategies and three alternative deep reinforcement learning algorithms.
Visible light communication provides high-speed indoor wireless access by combining illumination and communication functionalities into a single platform. However, conventional multiple access schemes exhibit limited capability in handling interference and improving spectral efficiency within dynamic visible light environments. To address these operational constraints, the researchers examined an indoor architecture that incorporates rate splitting multiple access to serve moving terrestrial users. Taking both communication quality and operational energy consumption into account, the authors formulated an optimization problem that jointly optimizes the precoding matrix and the portion allocation of the common stream to maximize system-wide energy efficiency.
The joint optimization problem presents significant mathematical difficulty because of coupled optimization variables and continuous dynamic characteristics in the operational indoor environment. To resolve these challenges, the research team designed an improved deep deterministic policy gradient algorithm featuring an integrated generative adversarial network. By incorporating the generative adversarial network specifically to train the critic network, the algorithm learns the precise distribution characteristics of the state-action value function. This architectural addition substantially improves the accuracy of state-action Q-value estimation during iterative reinforcement training.
Alongside the generative adversarial network, the researchers adopted a prioritized experience replay mechanism to strengthen the reinforcement learning procedure. This mechanism functions to improve sample efficiency and accelerate algorithmic convergence during continuous policy updates. Evaluated within dynamic simulated indoor environments, the resulting improved deep deterministic policy gradient algorithm consistently outperformed the two conventional strategies and three competing deep reinforcement learning algorithms, demonstrating enhanced energy efficiency while addressing the dual requirements of indoor lighting and wireless communication service.
Subject of Research: Energy efficiency optimization in dynamic indoor visible light communication
Article Title: Energy efficiency optimization for dynamic indoor VLC with RSMA via GenAI-based DRL
Article References: Tang, B., Yu, Y., Xie, W., & Chu, W. (2026). Energy efficiency optimization for dynamic indoor VLC with RSMA via GenAI-based DRL. Journal on Advances in Signal Processing. https://doi.org/10.1186/s13634-026-01379-z
Image Credits: AI Generated
DOI: 10.1186/s13634-026-01379-z
Keywords: visible light communication, indoor VLC, rate splitting multiple access, RSMA, energy efficiency, deep deterministic policy gradient, IDDPG, generative adversarial network, critic network, Q-value estimation, prioritized experience replay, sample efficiency, algorithm convergence, dynamic VLC, light-emitting diodes, precoding matrix, common stream allocation
Cite Scienmag News
Reid Dalton. (November 20, 2026). Generative AI Algorithm Boosts Energy Efficiency in Indoor Light Communication Simulations. Scienmag. https://scienmag.com/generative-ai-algorithm-boosts-energy-efficiency-in-indoor-light-communication-simulations/
Reid Dalton. "Generative AI Algorithm Boosts Energy Efficiency in Indoor Light Communication Simulations." Scienmag, 20 November 2026, https://scienmag.com/generative-ai-algorithm-boosts-energy-efficiency-in-indoor-light-communication-simulations/. Accessed 20 November 2026.
Reid Dalton. "Generative AI Algorithm Boosts Energy Efficiency in Indoor Light Communication Simulations." Scienmag. November 20, 2026. https://scienmag.com/generative-ai-algorithm-boosts-energy-efficiency-in-indoor-light-communication-simulations/








