Cooperative Multi-View Sensing with Hybrid Fusion and Channel-Aware Encoding for Environment-Aware 6G Networks

Abstract

The 6G network technology aims to achieve environment-aware intelligence through ISAC. The single-node-based sensing scheme has inherent limitations related to occlusion, angle, and interference, making the solution unreliable in mission-critical scenarios.  In this paper, we present the concept of Cooperative Multi-View Sensing (CMS), which leverages geographically scattered base stations, user equipment, and reconfigurable intelligent surfaces for creating a highly accurate map of the environment. The proposed CMS approach creatively merges three capabilities: (1) neural networks at edge devices are developed to detect semantic features from OFDM signals; (2) channel-aware adaptive encoding of communications lowers the latency by 42% at 15 dB due to compression based on the current signal-to-noise ratio; and (3) a hybrid fusion mechanism along with edge-level processing of local submaps using multi-view fusion networks. Results: The CMS obtains an 81.2% reduction in latency (293 ms to 55 ms) over the baselines and beats single-node sensing in terms of accuracy by 33%. An ablation study proves that adaptive encoding and hybrid fusion are equally important because removing either would result in up to 39-116% degradation in latency and 2-10% in accuracy. The scalability analysis proves O(n) scalability, in which the latency is 90 ms for 50 nodes, whereas the centralized (1290 ms) and distributed (622 ms) approaches become unmanageable. The statistical analysis based on 50 Monte Carlo runs validates the robustness of the approach (p < 0.001), with the optimal number of nodes being 8-12. CMS is the template for 6G wireless communication systems that are environmentally conscious and support real-time applications using collaborative multi-view sensors. Future implementations of 6G must ensure hybrid fusion with a number of 8 to 12 nodes and channel-aware encoding for adaptive compression.