Information-Theoretic Analysis of Vision-Aided ISAC over a Discrete Memoryless Channel

We investigate a vision-aided integrated sensing and communications (ISAC) system comprising a transmitter, a receiver and a vision sensor (such as a camera) co-located with the receiver. The vision-aided ISAC system uses the vision sensor to sense the environment and share the vision data with the receiver. The receiver decodes the transmitted message using the received signal and the vision data. Even though this vision data may not completely determine the channel impulse response, some information about the environment, such as whether the transmitter is visible from the receiver, could be potentially useful for decoding. The objective of this paper is to understand the value of such information, termed channel state knowledge, using an information-theoretic formalism. We examine three scenarios in which the vision sensor provides different amounts of channel state knowledge to the receiver: perfect, imperfect, and none. Further, we analyze the mutual information for the vision-aided ISAC system using joint and sequential processing approaches and demonstrate that the system with the joint processing of vision data and communication signals has higher mutual information. This analysis provides crucial insights into the performance limits of vision-aided ISAC systems.