Researchers propose event-triggered diffusion filter to reduce communication costs in multi-agent collaborative perception without sacrificing tracking accuracy

A team led by Jirong Zha submitted a paper describing an error-minimized event-triggered cubature information filter (CIF) for local estimation combined with a correlation-aware diffusion strategy for global fusion in multi-agent systems.

The proposed EDC-CIF algorithm addresses a known trade-off between estimation accuracy and communication cost by achieving improved tracking accuracy, reduced data transmission, and accelerated convergence simultaneously, according to experimental results presented in the paper.

The framework is designed for real-time multi-agent collaborative target tracking (the task of multiple AI agents working together to track moving objects) and demonstrates scalability across different system sizes.