Research project

Belief propagation-based target handover

A distributed integrated sensing and communication method that transfers target beliefs—and, when useful, associated measurements—between base stations with overlapping fields of view while avoiding repeated information transfer.

Method

What is handed over—and when

Criterion Existence × receiver visibility

A potential target is considered when its posterior existence exceeds the declaration threshold and its predicted state has sufficient probability of detection at the receiving base station.

Message Prior plus one associated measurement

The target prior is transferred once; the most likely associated measurement can also be shared. Local labels prevent repeated prior handovers and double counting.

Receiver Reuse the local BP tracker

The receiving base station appends transferred priors to its legacy targets, then processes local and handed-over measurements sequentially in its existing belief-propagation update.

Animation

Two-base-station simulation

Animated target handover tracking result between base stations
A two-base-station, two-target scenario with finite overlapping fields of view, missed detections, and clutter. The paper compares centralized processing, independent distributed tracking, and handover with or without measurement exchange over 100 Monte Carlo trials. In this simulated setting, handover with measurements achieves overall GOSPA comparable to centralized processing; the main gain is reduced missed detections, while exchanging measurements also improves localization and false-alarm behavior in the overlap region.

Evidence boundary

What the result establishes

Demonstrated Simulation performance

The published evaluation supports comparable aggregate GOSPA to its centralized baseline in the specified two-BS urban simulation.

Trade-off Threshold-dependent errors

The handover threshold balances missed detections against false alarms near the overlap boundary; the best behavior is not threshold-free.

Not yet shown Real-network deployment

The paper leaves heterogeneous networks, more complex sensing configurations, standard integration, and real-time deployment to future work.

Resources

Project resources