Vector Quantized Diffusion Model-Based Joint Source-Channel Coding for Task-Oriented Agent Communication

In this paper, a novel joint source-channel coding (JSCC) scheme based on the vector quantized diffusion model is proposed to enable task-oriented communication for multi-agent systems. To make the agent communication compatible with digital communication systems, the feature vectors encoded by the parameterized encoder are quantized using a learnable codebook. Furthermore, to improve the reliability of feature vector transmission, a discrete channel denoising diffusion model (DCDDM) is designed. The DCDDM can learn the distribution of the sender agent's feature vectors, and then utilize this learned knowledge to correct the receiver agent's received feature vector. Based on the corrected feature vectors, the parameterized decoder recovers task-related information for the receiver agent's subsequent decision making. Finally, the encoding process of the sender agent as well as the decoding process of the receiver agent are modelled as action policies. The multi-agent collaboration is modelled as a Markov decision process (MDP). The optimal encoder and decoder of the JSCC scheme are then obtained by solving the MDP using a reinforcement learning algorithm. Simulation results show that the proposed scheme can improve the robustness of multi-agent cooperation against channel noise with a reduction in the transmission overhead.