A Multi-Service Real-Time Resource Scheduling Optimization Method Based on the O-RAN Architecture

The Open Radio Access Network (O-RAN) has emerged as a key architecture to address diverse service demands and resource allocation challenges due to its openness and intelligence. However, traditional algorithms struggle to balance quality of service (QoS) and energy optimization in dynamic, high-dimensional environments. This paper proposes an O-RAN resource allocation method combining latent space modeling with the Actor-Critic framework. Offline latent space enhancement extracts features from historical data, while reinforcement learning optimizes resource allocation policies for efficient online decision-making. Experimental results show superior performance in cumulative rewards, QoS satisfaction, and energy optimization, with remarkable stability and adaptability under dynamic loads, providing an efficient and scalable solution for next-generation wireless networks.