Experimental Demonstration of Multi-Target Tracking in Integrated Sensing and Communication

TL;DR

This paper demonstrates multi-target tracking using a PHD filter in a real 5G ISAC setup within a factory, showing promising results despite real-world challenges.

Contribution

It is the first to apply multi-target tracking algorithms to real-world ISAC data, addressing practical challenges like clutter and hardware limitations.

Findings

Mean absolute ranging error <1.5m

Detection rates >91% in challenging scenarios

Successful end-to-end pipeline from measurement to evaluation

Abstract

For a wide range of envisioned integrated sensing and communication (ISAC) use cases, it is necessary to incorporate tracking techniques into cellular communication systems. While numerous multi-target tracking (MTT) algorithms exist, they have not yet been applied to real-world ISAC, with its challenges such as clutter and non-optimal hardware with design emphasis on communication instead of sensing. In this work, we showcase MTT based on the probability hypothesis density (PHD) filter in the range and radial speed domain. The measurements are taken with a 5G compliant ISAC proof-of-concept in a real factory environment, where the pedestrian-like targets are generated by a radar target emulator. We detail the complete pipeline, from measurement acquisition to evaluation, with a focus on the post-processing of the raw captured data and the tracking itself. Our end-to-end evaluation and comparison to simulations show good MTT performance with mean absolute ranging error <1.5m and detection rates >91% for realistic but challenging scenarios.