Deep Fusion of Multi-Object Densities Using Transformer

TL;DR

This paper introduces a transformer-based deep learning approach for fusing multi-object densities from multiple sensors, demonstrating superior performance over traditional Bayesian methods in simulated scenarios.

Contribution

The paper adapts a transformer architecture for multi-object density fusion, providing a novel deep learning method that improves over existing model-based Bayesian fusion techniques.

Findings

Transformer-based fusion outperforms Bayesian fusion in simulations

Deep learning effectively fuses multi-object densities from multiple sensors

Method shows robustness across different parameter settings

Abstract

In this paper, we demonstrate that deep learning based method can be used to fuse multi-object densities. Given a scenario with several sensors with possibly different field-of-views, tracking is performed locally in each sensor by a tracker, which produces random finite set multi-object densities. To fuse outputs from different trackers, we adapt a recently proposed transformer-based multi-object tracker, where the fusion result is a global multi-object density, describing the set of all alive objects at the current time. We compare the performance of the transformer-based fusion method with a well-performing model-based Bayesian fusion method in several simulated scenarios with different parameter settings using synthetic data. The simulation results show that the transformer-based fusion method outperforms the model-based Bayesian method in our experimental scenarios.