Demoing the RFRL Gym: A Reinforcement Learning Testbed for Wireless Communications

Radio Frequency Reinforcement Learning (RFRL) is anticipated to be a widely leveraged technology in the next generation of wireless communication systems, particularly 6G systems and next-gen military communications. To support education, research, and innovation in RFRL technologies, an open source simulation and analysis tool specifically for simulating wireless communications applications (both commercial and military) is under development that leverages the well-known OpenAI Gymnasium framework. In this demonstration, the current feature-complete functionalities of the RFRL Gym are showcased, particularly the ability to train and evaluate open-source Gymnasium-compatible RL algorithms against a series of representative user-defined wireless scenarios. In particular, scenarios representing dynamic spectrum access scenarios, as well as jamming/anti-jamming scenarios, will be demoed. Additionally, two simulation modes of the RFRL Gym will be demonstrated, namely a high-level abstracted gamified mode for researchers with minimal background in wireless communications and a low-level expert mode simulating real wireless signals, channels, and sensing for researchers with expertise in wireless communications concepts. The goal of this demonstration is two-fold. First, to showcase and solicit feedback on how RFRL Gym can be helpful to experts in the field to test and evaluate candidate RL algorithms for next-generation wireless systems. Second, to showcase and solicit feedback on how RFRL Gym can be helpful to develop a better understanding of RF and RL concepts for researchers with minimal expertise in the area.