Goal-oriented Resource Allocation for Collaborative Integrated Sensing and Communication

A team of researchers from L2S (Laboratoire des Signaux et Systèmes) and VNU-UET (Vietnam National University, University of Engineering and Technology) has published a new paper on arXiv titled 'Goal-oriented Resource Allocation for Collaborative Integrated Sensing and Communication (ISAC).' The work tackles a core challenge for future 6G and IoT networks: how to efficiently allocate limited network resources (like energy and bandwidth) when smart devices must simultaneously perform sensing tasks (like radar detection) and communicate data. The authors propose a novel, scalable framework that treats the network's ultimate goal—such as accurately classifying objects in an environment—as the primary optimization target.

The key innovation is the use of a theoretical metric called 'discriminant gain' as a direct proxy for classification performance. This allows them to formulate and solve cross-layer optimization problems. They developed two scheduling policies: an 'independent' policy with lower complexity and a 'joint' policy that exploits correlations between devices for better performance under strict communication constraints. Both policies are solved using successive convex approximation and outperform baseline methods in experiments with synthetic and realistic radar datasets. The joint policy shows particular strength when device sensing data is strongly correlated.