<p>To provide ubiquitous wireless connectivity for multiple Internet-of-Things devices, visible light communication (VLC) has become an alternative wireless solution to compensate for limitations in radiofrequency (RF) communication. As a result, the hybrid VLC/RF technology can be exploited in a complementary fashion to the existing indoor RF networks. In this study, we investigate the power and spectrum allocation problems for the VLC/RF system. Based on the multi-agent learning and cooperative game theory, we design two learning game models. To solve the power and spectrum allocation problems, we adopt the idea of Shapley value and bargaining solutions to estimate the rewards of multiple agents in the learning process. In the power allocation game, each access point (AP) dynamically adjusts its power levels by considering other APs’ decisions. In the spectrum allocation game, the spectrum resource of each AP is effectively shared by learning the most suitable bargaining solution. These two learning games are jointly combined, and work together to find the best compromise solution in the VLC/RF system platform. Therefore, our learning game approach can induce selfish agents to participate in a cooperative control process. To confirm the superiority of our proposed scheme, the numerical results demonstrate that our proposed scheme can improve the performance of VLC/RF system and device profits under dynamic network environments. Finally, concluding remarks and suggestions for future research are given. The findings of this study have significance for a wide range of future research fields.</p>